<\/p>\n
Values determined by this site’s code when the information was collected. Dates are JST.<\/em><\/p>\n<\/p>\nOverview<\/h2>\nggml-org has released “ggml-org\/OpenJev-GGUF”, a GGUF quantized version of “OpenJev”, an open-weight model specialized in decision-making. This model is designed to accept text, web pages, and screenshots as inputs to perform choice-based decision-making (decision-model). Instead of generating free text and parsing it for user-requested questions in JSON format, it features a mechanism that directly outputs probabilities or scores for each choice in a single forward pass.<\/p>\nSpecifications<\/h2>\nThe specifications of the original model openjev\/openjev<\/code> are as follows:\n– Parameters: 27.4B\n– Architecture: Qwen3_5ForConditionalGeneration (qwen3_5)\n– Context length: Up to 16,384 tokens<\/p>\nPerformance<\/h2>\nHere are the benchmark results for the original model openjev\/openjev<\/code> published by its creators. Since this is a GGUF quantized version, the following figures are measured values on the unquantized original model.<\/p>\nFirst, here is a comparison with other models across 10,000 text-based questions (using 34 public sources):<\/p>\n\n\n\n\nmodel<\/th>\naccuracy<\/th>\n<\/tr>\n<\/thead>\n\n\nJev (hosted API)<\/td>\n85.4% (8,540 of 10,000)<\/td>\n<\/tr>\n\nOpenJev<\/strong><\/td>\n84.0%<\/strong> (8,403 of 10,000)<\/td>\n<\/tr>\n\nthe same base model before tuning, same readout, its own calibration<\/td>\n80.4% (8,036 of 10,000)<\/td>\n<\/tr>\n\nNimble 9B (open)<\/td>\n75.7% (7,574 of 10,000)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nFrom this table, we can see that OpenJev achieves an accuracy of 84.0%, significantly outperforming the pre-tuning base model (80.4%) and the open model Nimble 9B (75.7%). Meanwhile, while it fell slightly short of the hosted API Jev (hosted API) at 85.4%, it comes very close with a small difference of 1.4 points.<\/p>\nNext is the breakdown of accuracy by category:<\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nkind of work<\/th>\nquestions<\/th>\nOpenJev<\/th>\nJev (hosted API)<\/th>\nbefore tuning<\/th>\n<\/tr>\n<\/thead>\n\n\nintent \/ routing \/ topic<\/td>\n1,218<\/td>\n92.8%<\/td>\n92.8%<\/td>\n93.2%<\/td>\n<\/tr>\n\nsentiment \/ stance<\/td>\n1,214<\/td>\n82.1%<\/td>\n81.5%<\/td>\n78.3%<\/td>\n<\/tr>\n\nspam \/ hate<\/td>\n695<\/td>\n80.4%<\/td>\n81.2%<\/td>\n80.1%<\/td>\n<\/tr>\n\nlegal<\/td>\n1,305<\/td>\n78.9%<\/td>\n78.8%<\/td>\n77.8%<\/td>\n<\/tr>\n\nethics \/ policy judgement<\/td>\n877<\/td>\n78.1%<\/td>\n75.8%<\/td>\n65.5%<\/td>\n<\/tr>\n\ncommonsense reasoning<\/td>\n2,260<\/td>\n85.8%<\/td>\n88.3%<\/td>\n80.3%<\/td>\n<\/tr>\n\nscience \/ facts \/ claims<\/td>\n1,558<\/td>\n82.5%<\/td>\n89.0%<\/td>\n81.0%<\/td>\n<\/tr>\n\nreading + language<\/td>\n873<\/td>\n89.2%<\/td>\n89.3%<\/td>\n83.3%<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nLooking at these category results, OpenJev shows high accuracy in tasks such as “intent \/ routing \/ topic” (92.8%) and “reading + language” (89.2%). It also scores higher than the hosted API Jev in “ethics \/ policy judgement” (78.1%) and “sentiment \/ stance” (82.1%). However, in fields like “science \/ facts \/ claims” (82.5%) and “commonsense reasoning” (85.8%), it falls slightly behind the hosted API Jev (89.0% and 88.3% respectively).<\/p>\nAdditionally, the comparison results before and after tuning on agent-based decision-making tasks are as follows:<\/p>\n\n\n\n\ntest<\/th>\nsteps<\/th>\nbefore tuning<\/th>\nOpenJev<\/th>\n<\/tr>\n<\/thead>\n\n\ndesktop: next action from a screenshot<\/td>\n2,000<\/td>\n76.5%<\/td>\n88.0%<\/strong><\/td>\n<\/tr>\n\nweb: next action on unseen websites<\/td>\n975<\/td>\n68.5%<\/td>\n87.4%<\/strong><\/td>\n<\/tr>\n\nweb: next action on unseen domains<\/td>\n1,000<\/td>\n65.7%<\/td>\n84.5%<\/strong><\/td>\n<\/tr>\n\nanswer flips when the options are shuffled<\/td>\n2,000<\/td>\n18.5%<\/td>\n2.3%<\/strong><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nFrom these results, OpenJev achieves a significant accuracy improvement of over 10 points compared to pre-tuning when determining next actions for desktop operations from screenshots or on unseen websites and domains. Furthermore, answer flips when options are shuffled decreased drastically from 18.5% to 2.3%, demonstrating extremely high stability as a decision-making model.<\/p>\nComparisons regarding multilingual support and long-context reading comprehension are as follows:<\/p>\n\n\n\n\ntest<\/th>\nquestions<\/th>\nbefore tuning<\/th>\nOpenJev<\/th>\n<\/tr>\n<\/thead>\n\n\ninference in German, French, Hindi, Chinese (XNLI)<\/td>\n240<\/td>\n72.5%<\/td>\n82.5%<\/strong><\/td>\n<\/tr>\n\nintent in German, French, Hindi, Japanese (MASSIVE)<\/td>\n240<\/td>\n80.4%<\/td>\n85.8%<\/strong><\/td>\n<\/tr>\n\nquestions about 2,600 to 8,500-token articles (QuALITY)<\/td>\n120<\/td>\n91.7%<\/td>\n91.7%<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nThis table indicates that accuracy has improved over pre-tuning in multilingual tasks involving German, French, Hindi, Chinese, and Japanese. On the other hand, in question tasks regarding long articles (QuALITY), the score remains unchanged at 91.7% before and after tuning, maintaining equivalent performance.<\/p>\nFinally, here is the number of completed tasks when executing end-to-end browser tasks (100 types of MiniWoB tasks, 1 trial each):<\/p>\n\n\n\n\nmodel<\/th>\ntasks completed<\/th>\n<\/tr>\n<\/thead>\n\n\nOpenJev<\/strong><\/td>\n39<\/strong><\/td>\n<\/tr>\n\nJev (hosted API)<\/td>\n39<\/td>\n<\/tr>\n\nthe same base model before tuning<\/td>\n38<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nThese results confirm that in end-to-end browser tasks, OpenJev completes 39 tasks, matching the hosted Jev API and showing a slight improvement from the 38 tasks completed before tuning.<\/p>\nStrengths and Use Cases<\/h2>\nThis model is a “decision model” that accepts text or images (screenshots) as inputs and makes optimal decisions from pre-defined choices. Unlike general chat bots that generate free text before parsing, it calculates probabilities directly from token scores at the initial output position, enabling fast and reliable processing.<\/p>\nBased on the performance and features of the original model openjev\/openjev<\/code>, it is highly suited for the following applications and tasks:<\/p>\n\n\nRouting and Triage<\/strong>:\n Determining the intent, topic, responsible team, priority, or language of user inquiry messages to automatically route them to the appropriate destination.<\/p>\n<\/li>\n\nModeration and Safety Judgment<\/strong>:\n Evaluating whether input text falls under harmful content, spam, hate speech, policy violations, or legal threats.<\/p>\n<\/li>\n\nLLM as a Judge<\/strong>:\n Objectively determining whether a generative AI model’s output is based on reliable evidence (presence of hallucinations), follows provided evaluation criteria (rubrics), or which of multiple answers is superior.<\/p>\n<\/li>\n\nBusiness Processes and Document Management<\/strong>:\n Automating routine judgment tasks such as approving, withholding, or rejecting invoices, determining the severity of system alerts, or classifying specific clause types in contracts.<\/p>\n<\/li>\n\nBrowser and Desktop Agents<\/strong>:\n Reading screen screenshots, HTML (DOM structure), or JSON data to determine the next element to interact with or action to execute (such as clicks or input). It can also judge whether a task is complete or blocked.<\/p>\n<\/li>\n\nScoring<\/strong>:\n Calculating ordered step-by-step evaluations accompanied by expected values or confidence levels.<\/p>\n<\/li>\n<\/ul>\nFurthermore, because this model allows dynamic specification of choices (labels) per request, task-specific pre-training or fixed labeling is unnecessary. It can process up to 52 choices in a single request during a single forward pass, and can simultaneously handle multiple questions (e.g., routing, sentiment analysis, urgency) in one request. A major strength is its support for multiple languages including Japanese (English, German, French, Hindi, Chinese, and Japanese).<\/p>\nHow It Differs from Similar Models<\/h2>\nIn a previous article introduced on our site, Decision-Making Specialized Model “OpenJev” Testing GGUF Model Released<\/a>, we covered “tinyopenjev-for-testing-gguf”, also released by ggml-org. However, that was an ultra-small test-only model (approx. 34M parameters) created solely to verify whether inference engines and loaders function properly, and its output content carried no meaning.<\/p>\nIn contrast, the newly released “OpenJev-GGUF” is a practical quantized model that retains the entire structure of the original model openjev\/openjev<\/code>. Unlike the test model, it can be used directly for actual decision-making tasks and agent inference purposes.<\/p>\n<\/p>\nHardware Requirements<\/h2>\nEstimated requirements (calculated by Local Model Watch)<\/strong> \u2014 27.4B parameters (taken from the base model openjev\/openjev)<\/p>\n\n\n\n\nYour VRAM<\/th>\nQuantization<\/th>\nFile size<\/th>\nEst. memory needed<\/th>\n<\/tr>\n<\/thead>\n\n\n24GB (RTX 4090 \/ 3090, etc.)<\/td>\nQ4_K_M<\/td>\n17.7GB<\/td>\n21.2GB<\/td>\n<\/tr>\n\n32GB (RTX 5090, etc.)<\/td>\nQ8_0<\/td>\n26.6GB<\/td>\n32.0GB<\/td>\n<\/tr>\n\n80GB class (A100 \/ H100)<\/td>\nBF16<\/td>\n50.1GB<\/td>\n60.1GB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nMemory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model’s authors. Compare with other models in our VRAM quick reference<\/a>. What the quantization names mean: glossary<\/a>.<\/em><\/p>\n<\/p>\n<\/p>\nCan You Run It Locally?<\/h2>\nRuns in Ollama, LM Studio and llama.cpp as-is.<\/strong><\/p>\nIt is distributed in GGUF, so no conversion is needed.<\/p>\nLicense \u2014 cc-by-nc-4.0<\/code> (Commercial use prohibited):<\/strong> Commercial use is prohibited<\/strong> (NC = NonCommercial). Internal business use can also count as commercial, so avoid this license for work use.<\/p>\nCompression:<\/strong> the Q4_K_M build measures 5.56 bits per weight \u2014 about 35% the size of the original 16-bit weights, calculated by this site from the actual file sizes.<\/p>\nCompiled by this site’s code from the published formats, converted builds we have found, and each engine’s own model registry. “Not found” means we have not seen such a build, not that none exists. License summaries are not legal advice \u2014 check the publisher’s original terms before relying on them.<\/em><\/p>\n<\/p>\n<\/p>\nOur Own Measurements<\/h2>\nMeasurements We Did Not Take<\/h3>\nThis model’s license (cc-by-nc-4.0<\/code>) does not permit commercial use. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<\/p>\n<\/p>\nRecent Models in the Same Size Class<\/h2>\nModels with <\/em>15\u201340B<\/em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site’s estimates; licenses are as stated on the model cards.<\/em><\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nModel<\/th>\nParameters<\/th>\nSmallest VRAM tier<\/th>\nLicense<\/th>\nOur article<\/th>\n<\/tr>\n<\/thead>\n\n\nQwen\/Qwen3.8-27B<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Vision-Language Model: 80GB+ VRAM<\/a> (2026-10-02)<\/td>\n<\/tr>\n\nCloudflare\/clef<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Structured Decision-Making Model: 80GB+ VRAM<\/a> (2026-10-01)<\/td>\n<\/tr>\n\nQwen\/Qwen3.8-27B<\/td>\n27.8B<\/td>\n8GB<\/td>\napache-2.0<\/td>\nQwen3.8-27B Vision-Language Model: Our Test Answers, 8GB+ VRAM<\/a> (2026-09-26)<\/td>\n<\/tr>\n\nbartowski\/vectionlabs_Salience-27B-R6-GGUF<\/td>\n27.8B<\/td>\n12GB<\/td>\napache-2.0<\/td>\nvectionlabs_Salience-27B-R6-GGUF: Our Test Answers, 12GB+ VRAM<\/a> (2026-09-15)<\/td>\n<\/tr>\n\nagentionai\/Signal-3.8-27B-GGUF<\/td>\n27.8B<\/td>\n16GB<\/td>\napache-2.0<\/td>\nSignal-3.8-27B-GGUF: Our Test Answers, 16GB+ VRAM<\/a> (2026-09-12)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/p>\nHow to Get It<\/h2>\nThis model is distributed in GGUF format and can be run using inference engines such as llama.cpp<\/code>.<\/p>\nWhen launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
ggml-org has released “ggml-org\/OpenJev-GGUF”, a GGUF quantized version of “OpenJev”, an open-weight model specialized in decision-making. This model is designed to accept text, web pages, and screenshots as inputs to perform choice-based decision-making (decision-model). Instead of generating free text and parsing it for user-requested questions in JSON format, it features a mechanism that directly outputs probabilities or scores for each choice in a single forward pass.<\/p>\n
The specifications of the original model openjev\/openjev<\/code> are as follows:\n– Parameters: 27.4B\n– Architecture: Qwen3_5ForConditionalGeneration (qwen3_5)\n– Context length: Up to 16,384 tokens<\/p>\nPerformance<\/h2>\nHere are the benchmark results for the original model openjev\/openjev<\/code> published by its creators. Since this is a GGUF quantized version, the following figures are measured values on the unquantized original model.<\/p>\nFirst, here is a comparison with other models across 10,000 text-based questions (using 34 public sources):<\/p>\n\n\n\n\nmodel<\/th>\naccuracy<\/th>\n<\/tr>\n<\/thead>\n\n\nJev (hosted API)<\/td>\n85.4% (8,540 of 10,000)<\/td>\n<\/tr>\n\nOpenJev<\/strong><\/td>\n84.0%<\/strong> (8,403 of 10,000)<\/td>\n<\/tr>\n\nthe same base model before tuning, same readout, its own calibration<\/td>\n80.4% (8,036 of 10,000)<\/td>\n<\/tr>\n\nNimble 9B (open)<\/td>\n75.7% (7,574 of 10,000)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nFrom this table, we can see that OpenJev achieves an accuracy of 84.0%, significantly outperforming the pre-tuning base model (80.4%) and the open model Nimble 9B (75.7%). Meanwhile, while it fell slightly short of the hosted API Jev (hosted API) at 85.4%, it comes very close with a small difference of 1.4 points.<\/p>\nNext is the breakdown of accuracy by category:<\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nkind of work<\/th>\nquestions<\/th>\nOpenJev<\/th>\nJev (hosted API)<\/th>\nbefore tuning<\/th>\n<\/tr>\n<\/thead>\n\n\nintent \/ routing \/ topic<\/td>\n1,218<\/td>\n92.8%<\/td>\n92.8%<\/td>\n93.2%<\/td>\n<\/tr>\n\nsentiment \/ stance<\/td>\n1,214<\/td>\n82.1%<\/td>\n81.5%<\/td>\n78.3%<\/td>\n<\/tr>\n\nspam \/ hate<\/td>\n695<\/td>\n80.4%<\/td>\n81.2%<\/td>\n80.1%<\/td>\n<\/tr>\n\nlegal<\/td>\n1,305<\/td>\n78.9%<\/td>\n78.8%<\/td>\n77.8%<\/td>\n<\/tr>\n\nethics \/ policy judgement<\/td>\n877<\/td>\n78.1%<\/td>\n75.8%<\/td>\n65.5%<\/td>\n<\/tr>\n\ncommonsense reasoning<\/td>\n2,260<\/td>\n85.8%<\/td>\n88.3%<\/td>\n80.3%<\/td>\n<\/tr>\n\nscience \/ facts \/ claims<\/td>\n1,558<\/td>\n82.5%<\/td>\n89.0%<\/td>\n81.0%<\/td>\n<\/tr>\n\nreading + language<\/td>\n873<\/td>\n89.2%<\/td>\n89.3%<\/td>\n83.3%<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nLooking at these category results, OpenJev shows high accuracy in tasks such as “intent \/ routing \/ topic” (92.8%) and “reading + language” (89.2%). It also scores higher than the hosted API Jev in “ethics \/ policy judgement” (78.1%) and “sentiment \/ stance” (82.1%). However, in fields like “science \/ facts \/ claims” (82.5%) and “commonsense reasoning” (85.8%), it falls slightly behind the hosted API Jev (89.0% and 88.3% respectively).<\/p>\nAdditionally, the comparison results before and after tuning on agent-based decision-making tasks are as follows:<\/p>\n\n\n\n\ntest<\/th>\nsteps<\/th>\nbefore tuning<\/th>\nOpenJev<\/th>\n<\/tr>\n<\/thead>\n\n\ndesktop: next action from a screenshot<\/td>\n2,000<\/td>\n76.5%<\/td>\n88.0%<\/strong><\/td>\n<\/tr>\n\nweb: next action on unseen websites<\/td>\n975<\/td>\n68.5%<\/td>\n87.4%<\/strong><\/td>\n<\/tr>\n\nweb: next action on unseen domains<\/td>\n1,000<\/td>\n65.7%<\/td>\n84.5%<\/strong><\/td>\n<\/tr>\n\nanswer flips when the options are shuffled<\/td>\n2,000<\/td>\n18.5%<\/td>\n2.3%<\/strong><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nFrom these results, OpenJev achieves a significant accuracy improvement of over 10 points compared to pre-tuning when determining next actions for desktop operations from screenshots or on unseen websites and domains. Furthermore, answer flips when options are shuffled decreased drastically from 18.5% to 2.3%, demonstrating extremely high stability as a decision-making model.<\/p>\nComparisons regarding multilingual support and long-context reading comprehension are as follows:<\/p>\n\n\n\n\ntest<\/th>\nquestions<\/th>\nbefore tuning<\/th>\nOpenJev<\/th>\n<\/tr>\n<\/thead>\n\n\ninference in German, French, Hindi, Chinese (XNLI)<\/td>\n240<\/td>\n72.5%<\/td>\n82.5%<\/strong><\/td>\n<\/tr>\n\nintent in German, French, Hindi, Japanese (MASSIVE)<\/td>\n240<\/td>\n80.4%<\/td>\n85.8%<\/strong><\/td>\n<\/tr>\n\nquestions about 2,600 to 8,500-token articles (QuALITY)<\/td>\n120<\/td>\n91.7%<\/td>\n91.7%<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nThis table indicates that accuracy has improved over pre-tuning in multilingual tasks involving German, French, Hindi, Chinese, and Japanese. On the other hand, in question tasks regarding long articles (QuALITY), the score remains unchanged at 91.7% before and after tuning, maintaining equivalent performance.<\/p>\nFinally, here is the number of completed tasks when executing end-to-end browser tasks (100 types of MiniWoB tasks, 1 trial each):<\/p>\n\n\n\n\nmodel<\/th>\ntasks completed<\/th>\n<\/tr>\n<\/thead>\n\n\nOpenJev<\/strong><\/td>\n39<\/strong><\/td>\n<\/tr>\n\nJev (hosted API)<\/td>\n39<\/td>\n<\/tr>\n\nthe same base model before tuning<\/td>\n38<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nThese results confirm that in end-to-end browser tasks, OpenJev completes 39 tasks, matching the hosted Jev API and showing a slight improvement from the 38 tasks completed before tuning.<\/p>\nStrengths and Use Cases<\/h2>\nThis model is a “decision model” that accepts text or images (screenshots) as inputs and makes optimal decisions from pre-defined choices. Unlike general chat bots that generate free text before parsing, it calculates probabilities directly from token scores at the initial output position, enabling fast and reliable processing.<\/p>\nBased on the performance and features of the original model openjev\/openjev<\/code>, it is highly suited for the following applications and tasks:<\/p>\n\n\nRouting and Triage<\/strong>:\n Determining the intent, topic, responsible team, priority, or language of user inquiry messages to automatically route them to the appropriate destination.<\/p>\n<\/li>\n\nModeration and Safety Judgment<\/strong>:\n Evaluating whether input text falls under harmful content, spam, hate speech, policy violations, or legal threats.<\/p>\n<\/li>\n\nLLM as a Judge<\/strong>:\n Objectively determining whether a generative AI model’s output is based on reliable evidence (presence of hallucinations), follows provided evaluation criteria (rubrics), or which of multiple answers is superior.<\/p>\n<\/li>\n\nBusiness Processes and Document Management<\/strong>:\n Automating routine judgment tasks such as approving, withholding, or rejecting invoices, determining the severity of system alerts, or classifying specific clause types in contracts.<\/p>\n<\/li>\n\nBrowser and Desktop Agents<\/strong>:\n Reading screen screenshots, HTML (DOM structure), or JSON data to determine the next element to interact with or action to execute (such as clicks or input). It can also judge whether a task is complete or blocked.<\/p>\n<\/li>\n\nScoring<\/strong>:\n Calculating ordered step-by-step evaluations accompanied by expected values or confidence levels.<\/p>\n<\/li>\n<\/ul>\nFurthermore, because this model allows dynamic specification of choices (labels) per request, task-specific pre-training or fixed labeling is unnecessary. It can process up to 52 choices in a single request during a single forward pass, and can simultaneously handle multiple questions (e.g., routing, sentiment analysis, urgency) in one request. A major strength is its support for multiple languages including Japanese (English, German, French, Hindi, Chinese, and Japanese).<\/p>\nHow It Differs from Similar Models<\/h2>\nIn a previous article introduced on our site, Decision-Making Specialized Model “OpenJev” Testing GGUF Model Released<\/a>, we covered “tinyopenjev-for-testing-gguf”, also released by ggml-org. However, that was an ultra-small test-only model (approx. 34M parameters) created solely to verify whether inference engines and loaders function properly, and its output content carried no meaning.<\/p>\nIn contrast, the newly released “OpenJev-GGUF” is a practical quantized model that retains the entire structure of the original model openjev\/openjev<\/code>. Unlike the test model, it can be used directly for actual decision-making tasks and agent inference purposes.<\/p>\n<\/p>\nHardware Requirements<\/h2>\nEstimated requirements (calculated by Local Model Watch)<\/strong> \u2014 27.4B parameters (taken from the base model openjev\/openjev)<\/p>\n\n\n\n\nYour VRAM<\/th>\nQuantization<\/th>\nFile size<\/th>\nEst. memory needed<\/th>\n<\/tr>\n<\/thead>\n\n\n24GB (RTX 4090 \/ 3090, etc.)<\/td>\nQ4_K_M<\/td>\n17.7GB<\/td>\n21.2GB<\/td>\n<\/tr>\n\n32GB (RTX 5090, etc.)<\/td>\nQ8_0<\/td>\n26.6GB<\/td>\n32.0GB<\/td>\n<\/tr>\n\n80GB class (A100 \/ H100)<\/td>\nBF16<\/td>\n50.1GB<\/td>\n60.1GB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nMemory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model’s authors. Compare with other models in our VRAM quick reference<\/a>. What the quantization names mean: glossary<\/a>.<\/em><\/p>\n<\/p>\n<\/p>\nCan You Run It Locally?<\/h2>\nRuns in Ollama, LM Studio and llama.cpp as-is.<\/strong><\/p>\nIt is distributed in GGUF, so no conversion is needed.<\/p>\nLicense \u2014 cc-by-nc-4.0<\/code> (Commercial use prohibited):<\/strong> Commercial use is prohibited<\/strong> (NC = NonCommercial). Internal business use can also count as commercial, so avoid this license for work use.<\/p>\nCompression:<\/strong> the Q4_K_M build measures 5.56 bits per weight \u2014 about 35% the size of the original 16-bit weights, calculated by this site from the actual file sizes.<\/p>\nCompiled by this site’s code from the published formats, converted builds we have found, and each engine’s own model registry. “Not found” means we have not seen such a build, not that none exists. License summaries are not legal advice \u2014 check the publisher’s original terms before relying on them.<\/em><\/p>\n<\/p>\n<\/p>\nOur Own Measurements<\/h2>\nMeasurements We Did Not Take<\/h3>\nThis model’s license (cc-by-nc-4.0<\/code>) does not permit commercial use. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<\/p>\n<\/p>\nRecent Models in the Same Size Class<\/h2>\nModels with <\/em>15\u201340B<\/em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site’s estimates; licenses are as stated on the model cards.<\/em><\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nModel<\/th>\nParameters<\/th>\nSmallest VRAM tier<\/th>\nLicense<\/th>\nOur article<\/th>\n<\/tr>\n<\/thead>\n\n\nQwen\/Qwen3.8-27B<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Vision-Language Model: 80GB+ VRAM<\/a> (2026-10-02)<\/td>\n<\/tr>\n\nCloudflare\/clef<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Structured Decision-Making Model: 80GB+ VRAM<\/a> (2026-10-01)<\/td>\n<\/tr>\n\nQwen\/Qwen3.8-27B<\/td>\n27.8B<\/td>\n8GB<\/td>\napache-2.0<\/td>\nQwen3.8-27B Vision-Language Model: Our Test Answers, 8GB+ VRAM<\/a> (2026-09-26)<\/td>\n<\/tr>\n\nbartowski\/vectionlabs_Salience-27B-R6-GGUF<\/td>\n27.8B<\/td>\n12GB<\/td>\napache-2.0<\/td>\nvectionlabs_Salience-27B-R6-GGUF: Our Test Answers, 12GB+ VRAM<\/a> (2026-09-15)<\/td>\n<\/tr>\n\nagentionai\/Signal-3.8-27B-GGUF<\/td>\n27.8B<\/td>\n16GB<\/td>\napache-2.0<\/td>\nSignal-3.8-27B-GGUF: Our Test Answers, 16GB+ VRAM<\/a> (2026-09-12)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/p>\nHow to Get It<\/h2>\nThis model is distributed in GGUF format and can be run using inference engines such as llama.cpp<\/code>.<\/p>\nWhen launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
openjev\/openjev<\/code> are as follows:\n– Parameters: 27.4B\n– Architecture: Qwen3_5ForConditionalGeneration (qwen3_5)\n– Context length: Up to 16,384 tokens<\/p>\nPerformance<\/h2>\nHere are the benchmark results for the original model openjev\/openjev<\/code> published by its creators. Since this is a GGUF quantized version, the following figures are measured values on the unquantized original model.<\/p>\nFirst, here is a comparison with other models across 10,000 text-based questions (using 34 public sources):<\/p>\n\n\n\n\nmodel<\/th>\naccuracy<\/th>\n<\/tr>\n<\/thead>\n\n\nJev (hosted API)<\/td>\n85.4% (8,540 of 10,000)<\/td>\n<\/tr>\n\nOpenJev<\/strong><\/td>\n84.0%<\/strong> (8,403 of 10,000)<\/td>\n<\/tr>\n\nthe same base model before tuning, same readout, its own calibration<\/td>\n80.4% (8,036 of 10,000)<\/td>\n<\/tr>\n\nNimble 9B (open)<\/td>\n75.7% (7,574 of 10,000)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nFrom this table, we can see that OpenJev achieves an accuracy of 84.0%, significantly outperforming the pre-tuning base model (80.4%) and the open model Nimble 9B (75.7%). Meanwhile, while it fell slightly short of the hosted API Jev (hosted API) at 85.4%, it comes very close with a small difference of 1.4 points.<\/p>\nNext is the breakdown of accuracy by category:<\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nkind of work<\/th>\nquestions<\/th>\nOpenJev<\/th>\nJev (hosted API)<\/th>\nbefore tuning<\/th>\n<\/tr>\n<\/thead>\n\n\nintent \/ routing \/ topic<\/td>\n1,218<\/td>\n92.8%<\/td>\n92.8%<\/td>\n93.2%<\/td>\n<\/tr>\n\nsentiment \/ stance<\/td>\n1,214<\/td>\n82.1%<\/td>\n81.5%<\/td>\n78.3%<\/td>\n<\/tr>\n\nspam \/ hate<\/td>\n695<\/td>\n80.4%<\/td>\n81.2%<\/td>\n80.1%<\/td>\n<\/tr>\n\nlegal<\/td>\n1,305<\/td>\n78.9%<\/td>\n78.8%<\/td>\n77.8%<\/td>\n<\/tr>\n\nethics \/ policy judgement<\/td>\n877<\/td>\n78.1%<\/td>\n75.8%<\/td>\n65.5%<\/td>\n<\/tr>\n\ncommonsense reasoning<\/td>\n2,260<\/td>\n85.8%<\/td>\n88.3%<\/td>\n80.3%<\/td>\n<\/tr>\n\nscience \/ facts \/ claims<\/td>\n1,558<\/td>\n82.5%<\/td>\n89.0%<\/td>\n81.0%<\/td>\n<\/tr>\n\nreading + language<\/td>\n873<\/td>\n89.2%<\/td>\n89.3%<\/td>\n83.3%<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nLooking at these category results, OpenJev shows high accuracy in tasks such as “intent \/ routing \/ topic” (92.8%) and “reading + language” (89.2%). It also scores higher than the hosted API Jev in “ethics \/ policy judgement” (78.1%) and “sentiment \/ stance” (82.1%). However, in fields like “science \/ facts \/ claims” (82.5%) and “commonsense reasoning” (85.8%), it falls slightly behind the hosted API Jev (89.0% and 88.3% respectively).<\/p>\nAdditionally, the comparison results before and after tuning on agent-based decision-making tasks are as follows:<\/p>\n\n\n\n\ntest<\/th>\nsteps<\/th>\nbefore tuning<\/th>\nOpenJev<\/th>\n<\/tr>\n<\/thead>\n\n\ndesktop: next action from a screenshot<\/td>\n2,000<\/td>\n76.5%<\/td>\n88.0%<\/strong><\/td>\n<\/tr>\n\nweb: next action on unseen websites<\/td>\n975<\/td>\n68.5%<\/td>\n87.4%<\/strong><\/td>\n<\/tr>\n\nweb: next action on unseen domains<\/td>\n1,000<\/td>\n65.7%<\/td>\n84.5%<\/strong><\/td>\n<\/tr>\n\nanswer flips when the options are shuffled<\/td>\n2,000<\/td>\n18.5%<\/td>\n2.3%<\/strong><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nFrom these results, OpenJev achieves a significant accuracy improvement of over 10 points compared to pre-tuning when determining next actions for desktop operations from screenshots or on unseen websites and domains. Furthermore, answer flips when options are shuffled decreased drastically from 18.5% to 2.3%, demonstrating extremely high stability as a decision-making model.<\/p>\nComparisons regarding multilingual support and long-context reading comprehension are as follows:<\/p>\n\n\n\n\ntest<\/th>\nquestions<\/th>\nbefore tuning<\/th>\nOpenJev<\/th>\n<\/tr>\n<\/thead>\n\n\ninference in German, French, Hindi, Chinese (XNLI)<\/td>\n240<\/td>\n72.5%<\/td>\n82.5%<\/strong><\/td>\n<\/tr>\n\nintent in German, French, Hindi, Japanese (MASSIVE)<\/td>\n240<\/td>\n80.4%<\/td>\n85.8%<\/strong><\/td>\n<\/tr>\n\nquestions about 2,600 to 8,500-token articles (QuALITY)<\/td>\n120<\/td>\n91.7%<\/td>\n91.7%<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nThis table indicates that accuracy has improved over pre-tuning in multilingual tasks involving German, French, Hindi, Chinese, and Japanese. On the other hand, in question tasks regarding long articles (QuALITY), the score remains unchanged at 91.7% before and after tuning, maintaining equivalent performance.<\/p>\nFinally, here is the number of completed tasks when executing end-to-end browser tasks (100 types of MiniWoB tasks, 1 trial each):<\/p>\n\n\n\n\nmodel<\/th>\ntasks completed<\/th>\n<\/tr>\n<\/thead>\n\n\nOpenJev<\/strong><\/td>\n39<\/strong><\/td>\n<\/tr>\n\nJev (hosted API)<\/td>\n39<\/td>\n<\/tr>\n\nthe same base model before tuning<\/td>\n38<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nThese results confirm that in end-to-end browser tasks, OpenJev completes 39 tasks, matching the hosted Jev API and showing a slight improvement from the 38 tasks completed before tuning.<\/p>\nStrengths and Use Cases<\/h2>\nThis model is a “decision model” that accepts text or images (screenshots) as inputs and makes optimal decisions from pre-defined choices. Unlike general chat bots that generate free text before parsing, it calculates probabilities directly from token scores at the initial output position, enabling fast and reliable processing.<\/p>\nBased on the performance and features of the original model openjev\/openjev<\/code>, it is highly suited for the following applications and tasks:<\/p>\n\n\nRouting and Triage<\/strong>:\n Determining the intent, topic, responsible team, priority, or language of user inquiry messages to automatically route them to the appropriate destination.<\/p>\n<\/li>\n\nModeration and Safety Judgment<\/strong>:\n Evaluating whether input text falls under harmful content, spam, hate speech, policy violations, or legal threats.<\/p>\n<\/li>\n\nLLM as a Judge<\/strong>:\n Objectively determining whether a generative AI model’s output is based on reliable evidence (presence of hallucinations), follows provided evaluation criteria (rubrics), or which of multiple answers is superior.<\/p>\n<\/li>\n\nBusiness Processes and Document Management<\/strong>:\n Automating routine judgment tasks such as approving, withholding, or rejecting invoices, determining the severity of system alerts, or classifying specific clause types in contracts.<\/p>\n<\/li>\n\nBrowser and Desktop Agents<\/strong>:\n Reading screen screenshots, HTML (DOM structure), or JSON data to determine the next element to interact with or action to execute (such as clicks or input). It can also judge whether a task is complete or blocked.<\/p>\n<\/li>\n\nScoring<\/strong>:\n Calculating ordered step-by-step evaluations accompanied by expected values or confidence levels.<\/p>\n<\/li>\n<\/ul>\nFurthermore, because this model allows dynamic specification of choices (labels) per request, task-specific pre-training or fixed labeling is unnecessary. It can process up to 52 choices in a single request during a single forward pass, and can simultaneously handle multiple questions (e.g., routing, sentiment analysis, urgency) in one request. A major strength is its support for multiple languages including Japanese (English, German, French, Hindi, Chinese, and Japanese).<\/p>\nHow It Differs from Similar Models<\/h2>\nIn a previous article introduced on our site, Decision-Making Specialized Model “OpenJev” Testing GGUF Model Released<\/a>, we covered “tinyopenjev-for-testing-gguf”, also released by ggml-org. However, that was an ultra-small test-only model (approx. 34M parameters) created solely to verify whether inference engines and loaders function properly, and its output content carried no meaning.<\/p>\nIn contrast, the newly released “OpenJev-GGUF” is a practical quantized model that retains the entire structure of the original model openjev\/openjev<\/code>. Unlike the test model, it can be used directly for actual decision-making tasks and agent inference purposes.<\/p>\n<\/p>\nHardware Requirements<\/h2>\nEstimated requirements (calculated by Local Model Watch)<\/strong> \u2014 27.4B parameters (taken from the base model openjev\/openjev)<\/p>\n\n\n\n\nYour VRAM<\/th>\nQuantization<\/th>\nFile size<\/th>\nEst. memory needed<\/th>\n<\/tr>\n<\/thead>\n\n\n24GB (RTX 4090 \/ 3090, etc.)<\/td>\nQ4_K_M<\/td>\n17.7GB<\/td>\n21.2GB<\/td>\n<\/tr>\n\n32GB (RTX 5090, etc.)<\/td>\nQ8_0<\/td>\n26.6GB<\/td>\n32.0GB<\/td>\n<\/tr>\n\n80GB class (A100 \/ H100)<\/td>\nBF16<\/td>\n50.1GB<\/td>\n60.1GB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nMemory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model’s authors. Compare with other models in our VRAM quick reference<\/a>. What the quantization names mean: glossary<\/a>.<\/em><\/p>\n<\/p>\n<\/p>\nCan You Run It Locally?<\/h2>\nRuns in Ollama, LM Studio and llama.cpp as-is.<\/strong><\/p>\nIt is distributed in GGUF, so no conversion is needed.<\/p>\nLicense \u2014 cc-by-nc-4.0<\/code> (Commercial use prohibited):<\/strong> Commercial use is prohibited<\/strong> (NC = NonCommercial). Internal business use can also count as commercial, so avoid this license for work use.<\/p>\nCompression:<\/strong> the Q4_K_M build measures 5.56 bits per weight \u2014 about 35% the size of the original 16-bit weights, calculated by this site from the actual file sizes.<\/p>\nCompiled by this site’s code from the published formats, converted builds we have found, and each engine’s own model registry. “Not found” means we have not seen such a build, not that none exists. License summaries are not legal advice \u2014 check the publisher’s original terms before relying on them.<\/em><\/p>\n<\/p>\n<\/p>\nOur Own Measurements<\/h2>\nMeasurements We Did Not Take<\/h3>\nThis model’s license (cc-by-nc-4.0<\/code>) does not permit commercial use. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<\/p>\n<\/p>\nRecent Models in the Same Size Class<\/h2>\nModels with <\/em>15\u201340B<\/em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site’s estimates; licenses are as stated on the model cards.<\/em><\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nModel<\/th>\nParameters<\/th>\nSmallest VRAM tier<\/th>\nLicense<\/th>\nOur article<\/th>\n<\/tr>\n<\/thead>\n\n\nQwen\/Qwen3.8-27B<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Vision-Language Model: 80GB+ VRAM<\/a> (2026-10-02)<\/td>\n<\/tr>\n\nCloudflare\/clef<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Structured Decision-Making Model: 80GB+ VRAM<\/a> (2026-10-01)<\/td>\n<\/tr>\n\nQwen\/Qwen3.8-27B<\/td>\n27.8B<\/td>\n8GB<\/td>\napache-2.0<\/td>\nQwen3.8-27B Vision-Language Model: Our Test Answers, 8GB+ VRAM<\/a> (2026-09-26)<\/td>\n<\/tr>\n\nbartowski\/vectionlabs_Salience-27B-R6-GGUF<\/td>\n27.8B<\/td>\n12GB<\/td>\napache-2.0<\/td>\nvectionlabs_Salience-27B-R6-GGUF: Our Test Answers, 12GB+ VRAM<\/a> (2026-09-15)<\/td>\n<\/tr>\n\nagentionai\/Signal-3.8-27B-GGUF<\/td>\n27.8B<\/td>\n16GB<\/td>\napache-2.0<\/td>\nSignal-3.8-27B-GGUF: Our Test Answers, 16GB+ VRAM<\/a> (2026-09-12)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/p>\nHow to Get It<\/h2>\nThis model is distributed in GGUF format and can be run using inference engines such as llama.cpp<\/code>.<\/p>\nWhen launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
Here are the benchmark results for the original model openjev\/openjev<\/code> published by its creators. Since this is a GGUF quantized version, the following figures are measured values on the unquantized original model.<\/p>\nFirst, here is a comparison with other models across 10,000 text-based questions (using 34 public sources):<\/p>\n\n\n\n\nmodel<\/th>\naccuracy<\/th>\n<\/tr>\n<\/thead>\n\n\nJev (hosted API)<\/td>\n85.4% (8,540 of 10,000)<\/td>\n<\/tr>\n\nOpenJev<\/strong><\/td>\n84.0%<\/strong> (8,403 of 10,000)<\/td>\n<\/tr>\n\nthe same base model before tuning, same readout, its own calibration<\/td>\n80.4% (8,036 of 10,000)<\/td>\n<\/tr>\n\nNimble 9B (open)<\/td>\n75.7% (7,574 of 10,000)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nFrom this table, we can see that OpenJev achieves an accuracy of 84.0%, significantly outperforming the pre-tuning base model (80.4%) and the open model Nimble 9B (75.7%). Meanwhile, while it fell slightly short of the hosted API Jev (hosted API) at 85.4%, it comes very close with a small difference of 1.4 points.<\/p>\nNext is the breakdown of accuracy by category:<\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nkind of work<\/th>\nquestions<\/th>\nOpenJev<\/th>\nJev (hosted API)<\/th>\nbefore tuning<\/th>\n<\/tr>\n<\/thead>\n\n\nintent \/ routing \/ topic<\/td>\n1,218<\/td>\n92.8%<\/td>\n92.8%<\/td>\n93.2%<\/td>\n<\/tr>\n\nsentiment \/ stance<\/td>\n1,214<\/td>\n82.1%<\/td>\n81.5%<\/td>\n78.3%<\/td>\n<\/tr>\n\nspam \/ hate<\/td>\n695<\/td>\n80.4%<\/td>\n81.2%<\/td>\n80.1%<\/td>\n<\/tr>\n\nlegal<\/td>\n1,305<\/td>\n78.9%<\/td>\n78.8%<\/td>\n77.8%<\/td>\n<\/tr>\n\nethics \/ policy judgement<\/td>\n877<\/td>\n78.1%<\/td>\n75.8%<\/td>\n65.5%<\/td>\n<\/tr>\n\ncommonsense reasoning<\/td>\n2,260<\/td>\n85.8%<\/td>\n88.3%<\/td>\n80.3%<\/td>\n<\/tr>\n\nscience \/ facts \/ claims<\/td>\n1,558<\/td>\n82.5%<\/td>\n89.0%<\/td>\n81.0%<\/td>\n<\/tr>\n\nreading + language<\/td>\n873<\/td>\n89.2%<\/td>\n89.3%<\/td>\n83.3%<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nLooking at these category results, OpenJev shows high accuracy in tasks such as “intent \/ routing \/ topic” (92.8%) and “reading + language” (89.2%). It also scores higher than the hosted API Jev in “ethics \/ policy judgement” (78.1%) and “sentiment \/ stance” (82.1%). However, in fields like “science \/ facts \/ claims” (82.5%) and “commonsense reasoning” (85.8%), it falls slightly behind the hosted API Jev (89.0% and 88.3% respectively).<\/p>\nAdditionally, the comparison results before and after tuning on agent-based decision-making tasks are as follows:<\/p>\n\n\n\n\ntest<\/th>\nsteps<\/th>\nbefore tuning<\/th>\nOpenJev<\/th>\n<\/tr>\n<\/thead>\n\n\ndesktop: next action from a screenshot<\/td>\n2,000<\/td>\n76.5%<\/td>\n88.0%<\/strong><\/td>\n<\/tr>\n\nweb: next action on unseen websites<\/td>\n975<\/td>\n68.5%<\/td>\n87.4%<\/strong><\/td>\n<\/tr>\n\nweb: next action on unseen domains<\/td>\n1,000<\/td>\n65.7%<\/td>\n84.5%<\/strong><\/td>\n<\/tr>\n\nanswer flips when the options are shuffled<\/td>\n2,000<\/td>\n18.5%<\/td>\n2.3%<\/strong><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nFrom these results, OpenJev achieves a significant accuracy improvement of over 10 points compared to pre-tuning when determining next actions for desktop operations from screenshots or on unseen websites and domains. Furthermore, answer flips when options are shuffled decreased drastically from 18.5% to 2.3%, demonstrating extremely high stability as a decision-making model.<\/p>\nComparisons regarding multilingual support and long-context reading comprehension are as follows:<\/p>\n\n\n\n\ntest<\/th>\nquestions<\/th>\nbefore tuning<\/th>\nOpenJev<\/th>\n<\/tr>\n<\/thead>\n\n\ninference in German, French, Hindi, Chinese (XNLI)<\/td>\n240<\/td>\n72.5%<\/td>\n82.5%<\/strong><\/td>\n<\/tr>\n\nintent in German, French, Hindi, Japanese (MASSIVE)<\/td>\n240<\/td>\n80.4%<\/td>\n85.8%<\/strong><\/td>\n<\/tr>\n\nquestions about 2,600 to 8,500-token articles (QuALITY)<\/td>\n120<\/td>\n91.7%<\/td>\n91.7%<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nThis table indicates that accuracy has improved over pre-tuning in multilingual tasks involving German, French, Hindi, Chinese, and Japanese. On the other hand, in question tasks regarding long articles (QuALITY), the score remains unchanged at 91.7% before and after tuning, maintaining equivalent performance.<\/p>\nFinally, here is the number of completed tasks when executing end-to-end browser tasks (100 types of MiniWoB tasks, 1 trial each):<\/p>\n\n\n\n\nmodel<\/th>\ntasks completed<\/th>\n<\/tr>\n<\/thead>\n\n\nOpenJev<\/strong><\/td>\n39<\/strong><\/td>\n<\/tr>\n\nJev (hosted API)<\/td>\n39<\/td>\n<\/tr>\n\nthe same base model before tuning<\/td>\n38<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nThese results confirm that in end-to-end browser tasks, OpenJev completes 39 tasks, matching the hosted Jev API and showing a slight improvement from the 38 tasks completed before tuning.<\/p>\nStrengths and Use Cases<\/h2>\nThis model is a “decision model” that accepts text or images (screenshots) as inputs and makes optimal decisions from pre-defined choices. Unlike general chat bots that generate free text before parsing, it calculates probabilities directly from token scores at the initial output position, enabling fast and reliable processing.<\/p>\nBased on the performance and features of the original model openjev\/openjev<\/code>, it is highly suited for the following applications and tasks:<\/p>\n\n\nRouting and Triage<\/strong>:\n Determining the intent, topic, responsible team, priority, or language of user inquiry messages to automatically route them to the appropriate destination.<\/p>\n<\/li>\n\nModeration and Safety Judgment<\/strong>:\n Evaluating whether input text falls under harmful content, spam, hate speech, policy violations, or legal threats.<\/p>\n<\/li>\n\nLLM as a Judge<\/strong>:\n Objectively determining whether a generative AI model’s output is based on reliable evidence (presence of hallucinations), follows provided evaluation criteria (rubrics), or which of multiple answers is superior.<\/p>\n<\/li>\n\nBusiness Processes and Document Management<\/strong>:\n Automating routine judgment tasks such as approving, withholding, or rejecting invoices, determining the severity of system alerts, or classifying specific clause types in contracts.<\/p>\n<\/li>\n\nBrowser and Desktop Agents<\/strong>:\n Reading screen screenshots, HTML (DOM structure), or JSON data to determine the next element to interact with or action to execute (such as clicks or input). It can also judge whether a task is complete or blocked.<\/p>\n<\/li>\n\nScoring<\/strong>:\n Calculating ordered step-by-step evaluations accompanied by expected values or confidence levels.<\/p>\n<\/li>\n<\/ul>\nFurthermore, because this model allows dynamic specification of choices (labels) per request, task-specific pre-training or fixed labeling is unnecessary. It can process up to 52 choices in a single request during a single forward pass, and can simultaneously handle multiple questions (e.g., routing, sentiment analysis, urgency) in one request. A major strength is its support for multiple languages including Japanese (English, German, French, Hindi, Chinese, and Japanese).<\/p>\nHow It Differs from Similar Models<\/h2>\nIn a previous article introduced on our site, Decision-Making Specialized Model “OpenJev” Testing GGUF Model Released<\/a>, we covered “tinyopenjev-for-testing-gguf”, also released by ggml-org. However, that was an ultra-small test-only model (approx. 34M parameters) created solely to verify whether inference engines and loaders function properly, and its output content carried no meaning.<\/p>\nIn contrast, the newly released “OpenJev-GGUF” is a practical quantized model that retains the entire structure of the original model openjev\/openjev<\/code>. Unlike the test model, it can be used directly for actual decision-making tasks and agent inference purposes.<\/p>\n<\/p>\nHardware Requirements<\/h2>\nEstimated requirements (calculated by Local Model Watch)<\/strong> \u2014 27.4B parameters (taken from the base model openjev\/openjev)<\/p>\n\n\n\n\nYour VRAM<\/th>\nQuantization<\/th>\nFile size<\/th>\nEst. memory needed<\/th>\n<\/tr>\n<\/thead>\n\n\n24GB (RTX 4090 \/ 3090, etc.)<\/td>\nQ4_K_M<\/td>\n17.7GB<\/td>\n21.2GB<\/td>\n<\/tr>\n\n32GB (RTX 5090, etc.)<\/td>\nQ8_0<\/td>\n26.6GB<\/td>\n32.0GB<\/td>\n<\/tr>\n\n80GB class (A100 \/ H100)<\/td>\nBF16<\/td>\n50.1GB<\/td>\n60.1GB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nMemory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model’s authors. Compare with other models in our VRAM quick reference<\/a>. What the quantization names mean: glossary<\/a>.<\/em><\/p>\n<\/p>\n<\/p>\nCan You Run It Locally?<\/h2>\nRuns in Ollama, LM Studio and llama.cpp as-is.<\/strong><\/p>\nIt is distributed in GGUF, so no conversion is needed.<\/p>\nLicense \u2014 cc-by-nc-4.0<\/code> (Commercial use prohibited):<\/strong> Commercial use is prohibited<\/strong> (NC = NonCommercial). Internal business use can also count as commercial, so avoid this license for work use.<\/p>\nCompression:<\/strong> the Q4_K_M build measures 5.56 bits per weight \u2014 about 35% the size of the original 16-bit weights, calculated by this site from the actual file sizes.<\/p>\nCompiled by this site’s code from the published formats, converted builds we have found, and each engine’s own model registry. “Not found” means we have not seen such a build, not that none exists. License summaries are not legal advice \u2014 check the publisher’s original terms before relying on them.<\/em><\/p>\n<\/p>\n<\/p>\nOur Own Measurements<\/h2>\nMeasurements We Did Not Take<\/h3>\nThis model’s license (cc-by-nc-4.0<\/code>) does not permit commercial use. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<\/p>\n<\/p>\nRecent Models in the Same Size Class<\/h2>\nModels with <\/em>15\u201340B<\/em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site’s estimates; licenses are as stated on the model cards.<\/em><\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nModel<\/th>\nParameters<\/th>\nSmallest VRAM tier<\/th>\nLicense<\/th>\nOur article<\/th>\n<\/tr>\n<\/thead>\n\n\nQwen\/Qwen3.8-27B<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Vision-Language Model: 80GB+ VRAM<\/a> (2026-10-02)<\/td>\n<\/tr>\n\nCloudflare\/clef<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Structured Decision-Making Model: 80GB+ VRAM<\/a> (2026-10-01)<\/td>\n<\/tr>\n\nQwen\/Qwen3.8-27B<\/td>\n27.8B<\/td>\n8GB<\/td>\napache-2.0<\/td>\nQwen3.8-27B Vision-Language Model: Our Test Answers, 8GB+ VRAM<\/a> (2026-09-26)<\/td>\n<\/tr>\n\nbartowski\/vectionlabs_Salience-27B-R6-GGUF<\/td>\n27.8B<\/td>\n12GB<\/td>\napache-2.0<\/td>\nvectionlabs_Salience-27B-R6-GGUF: Our Test Answers, 12GB+ VRAM<\/a> (2026-09-15)<\/td>\n<\/tr>\n\nagentionai\/Signal-3.8-27B-GGUF<\/td>\n27.8B<\/td>\n16GB<\/td>\napache-2.0<\/td>\nSignal-3.8-27B-GGUF: Our Test Answers, 16GB+ VRAM<\/a> (2026-09-12)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/p>\nHow to Get It<\/h2>\nThis model is distributed in GGUF format and can be run using inference engines such as llama.cpp<\/code>.<\/p>\nWhen launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
openjev\/openjev<\/code> published by its creators. Since this is a GGUF quantized version, the following figures are measured values on the unquantized original model.<\/p>\nFirst, here is a comparison with other models across 10,000 text-based questions (using 34 public sources):<\/p>\n\n\n\n\nmodel<\/th>\naccuracy<\/th>\n<\/tr>\n<\/thead>\n\n\nJev (hosted API)<\/td>\n85.4% (8,540 of 10,000)<\/td>\n<\/tr>\n\nOpenJev<\/strong><\/td>\n84.0%<\/strong> (8,403 of 10,000)<\/td>\n<\/tr>\n\nthe same base model before tuning, same readout, its own calibration<\/td>\n80.4% (8,036 of 10,000)<\/td>\n<\/tr>\n\nNimble 9B (open)<\/td>\n75.7% (7,574 of 10,000)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nFrom this table, we can see that OpenJev achieves an accuracy of 84.0%, significantly outperforming the pre-tuning base model (80.4%) and the open model Nimble 9B (75.7%). Meanwhile, while it fell slightly short of the hosted API Jev (hosted API) at 85.4%, it comes very close with a small difference of 1.4 points.<\/p>\nNext is the breakdown of accuracy by category:<\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nkind of work<\/th>\nquestions<\/th>\nOpenJev<\/th>\nJev (hosted API)<\/th>\nbefore tuning<\/th>\n<\/tr>\n<\/thead>\n\n\nintent \/ routing \/ topic<\/td>\n1,218<\/td>\n92.8%<\/td>\n92.8%<\/td>\n93.2%<\/td>\n<\/tr>\n\nsentiment \/ stance<\/td>\n1,214<\/td>\n82.1%<\/td>\n81.5%<\/td>\n78.3%<\/td>\n<\/tr>\n\nspam \/ hate<\/td>\n695<\/td>\n80.4%<\/td>\n81.2%<\/td>\n80.1%<\/td>\n<\/tr>\n\nlegal<\/td>\n1,305<\/td>\n78.9%<\/td>\n78.8%<\/td>\n77.8%<\/td>\n<\/tr>\n\nethics \/ policy judgement<\/td>\n877<\/td>\n78.1%<\/td>\n75.8%<\/td>\n65.5%<\/td>\n<\/tr>\n\ncommonsense reasoning<\/td>\n2,260<\/td>\n85.8%<\/td>\n88.3%<\/td>\n80.3%<\/td>\n<\/tr>\n\nscience \/ facts \/ claims<\/td>\n1,558<\/td>\n82.5%<\/td>\n89.0%<\/td>\n81.0%<\/td>\n<\/tr>\n\nreading + language<\/td>\n873<\/td>\n89.2%<\/td>\n89.3%<\/td>\n83.3%<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nLooking at these category results, OpenJev shows high accuracy in tasks such as “intent \/ routing \/ topic” (92.8%) and “reading + language” (89.2%). It also scores higher than the hosted API Jev in “ethics \/ policy judgement” (78.1%) and “sentiment \/ stance” (82.1%). However, in fields like “science \/ facts \/ claims” (82.5%) and “commonsense reasoning” (85.8%), it falls slightly behind the hosted API Jev (89.0% and 88.3% respectively).<\/p>\nAdditionally, the comparison results before and after tuning on agent-based decision-making tasks are as follows:<\/p>\n\n\n\n\ntest<\/th>\nsteps<\/th>\nbefore tuning<\/th>\nOpenJev<\/th>\n<\/tr>\n<\/thead>\n\n\ndesktop: next action from a screenshot<\/td>\n2,000<\/td>\n76.5%<\/td>\n88.0%<\/strong><\/td>\n<\/tr>\n\nweb: next action on unseen websites<\/td>\n975<\/td>\n68.5%<\/td>\n87.4%<\/strong><\/td>\n<\/tr>\n\nweb: next action on unseen domains<\/td>\n1,000<\/td>\n65.7%<\/td>\n84.5%<\/strong><\/td>\n<\/tr>\n\nanswer flips when the options are shuffled<\/td>\n2,000<\/td>\n18.5%<\/td>\n2.3%<\/strong><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nFrom these results, OpenJev achieves a significant accuracy improvement of over 10 points compared to pre-tuning when determining next actions for desktop operations from screenshots or on unseen websites and domains. Furthermore, answer flips when options are shuffled decreased drastically from 18.5% to 2.3%, demonstrating extremely high stability as a decision-making model.<\/p>\nComparisons regarding multilingual support and long-context reading comprehension are as follows:<\/p>\n\n\n\n\ntest<\/th>\nquestions<\/th>\nbefore tuning<\/th>\nOpenJev<\/th>\n<\/tr>\n<\/thead>\n\n\ninference in German, French, Hindi, Chinese (XNLI)<\/td>\n240<\/td>\n72.5%<\/td>\n82.5%<\/strong><\/td>\n<\/tr>\n\nintent in German, French, Hindi, Japanese (MASSIVE)<\/td>\n240<\/td>\n80.4%<\/td>\n85.8%<\/strong><\/td>\n<\/tr>\n\nquestions about 2,600 to 8,500-token articles (QuALITY)<\/td>\n120<\/td>\n91.7%<\/td>\n91.7%<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nThis table indicates that accuracy has improved over pre-tuning in multilingual tasks involving German, French, Hindi, Chinese, and Japanese. On the other hand, in question tasks regarding long articles (QuALITY), the score remains unchanged at 91.7% before and after tuning, maintaining equivalent performance.<\/p>\nFinally, here is the number of completed tasks when executing end-to-end browser tasks (100 types of MiniWoB tasks, 1 trial each):<\/p>\n\n\n\n\nmodel<\/th>\ntasks completed<\/th>\n<\/tr>\n<\/thead>\n\n\nOpenJev<\/strong><\/td>\n39<\/strong><\/td>\n<\/tr>\n\nJev (hosted API)<\/td>\n39<\/td>\n<\/tr>\n\nthe same base model before tuning<\/td>\n38<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nThese results confirm that in end-to-end browser tasks, OpenJev completes 39 tasks, matching the hosted Jev API and showing a slight improvement from the 38 tasks completed before tuning.<\/p>\nStrengths and Use Cases<\/h2>\nThis model is a “decision model” that accepts text or images (screenshots) as inputs and makes optimal decisions from pre-defined choices. Unlike general chat bots that generate free text before parsing, it calculates probabilities directly from token scores at the initial output position, enabling fast and reliable processing.<\/p>\nBased on the performance and features of the original model openjev\/openjev<\/code>, it is highly suited for the following applications and tasks:<\/p>\n\n\nRouting and Triage<\/strong>:\n Determining the intent, topic, responsible team, priority, or language of user inquiry messages to automatically route them to the appropriate destination.<\/p>\n<\/li>\n\nModeration and Safety Judgment<\/strong>:\n Evaluating whether input text falls under harmful content, spam, hate speech, policy violations, or legal threats.<\/p>\n<\/li>\n\nLLM as a Judge<\/strong>:\n Objectively determining whether a generative AI model’s output is based on reliable evidence (presence of hallucinations), follows provided evaluation criteria (rubrics), or which of multiple answers is superior.<\/p>\n<\/li>\n\nBusiness Processes and Document Management<\/strong>:\n Automating routine judgment tasks such as approving, withholding, or rejecting invoices, determining the severity of system alerts, or classifying specific clause types in contracts.<\/p>\n<\/li>\n\nBrowser and Desktop Agents<\/strong>:\n Reading screen screenshots, HTML (DOM structure), or JSON data to determine the next element to interact with or action to execute (such as clicks or input). It can also judge whether a task is complete or blocked.<\/p>\n<\/li>\n\nScoring<\/strong>:\n Calculating ordered step-by-step evaluations accompanied by expected values or confidence levels.<\/p>\n<\/li>\n<\/ul>\nFurthermore, because this model allows dynamic specification of choices (labels) per request, task-specific pre-training or fixed labeling is unnecessary. It can process up to 52 choices in a single request during a single forward pass, and can simultaneously handle multiple questions (e.g., routing, sentiment analysis, urgency) in one request. A major strength is its support for multiple languages including Japanese (English, German, French, Hindi, Chinese, and Japanese).<\/p>\nHow It Differs from Similar Models<\/h2>\nIn a previous article introduced on our site, Decision-Making Specialized Model “OpenJev” Testing GGUF Model Released<\/a>, we covered “tinyopenjev-for-testing-gguf”, also released by ggml-org. However, that was an ultra-small test-only model (approx. 34M parameters) created solely to verify whether inference engines and loaders function properly, and its output content carried no meaning.<\/p>\nIn contrast, the newly released “OpenJev-GGUF” is a practical quantized model that retains the entire structure of the original model openjev\/openjev<\/code>. Unlike the test model, it can be used directly for actual decision-making tasks and agent inference purposes.<\/p>\n<\/p>\nHardware Requirements<\/h2>\nEstimated requirements (calculated by Local Model Watch)<\/strong> \u2014 27.4B parameters (taken from the base model openjev\/openjev)<\/p>\n\n\n\n\nYour VRAM<\/th>\nQuantization<\/th>\nFile size<\/th>\nEst. memory needed<\/th>\n<\/tr>\n<\/thead>\n\n\n24GB (RTX 4090 \/ 3090, etc.)<\/td>\nQ4_K_M<\/td>\n17.7GB<\/td>\n21.2GB<\/td>\n<\/tr>\n\n32GB (RTX 5090, etc.)<\/td>\nQ8_0<\/td>\n26.6GB<\/td>\n32.0GB<\/td>\n<\/tr>\n\n80GB class (A100 \/ H100)<\/td>\nBF16<\/td>\n50.1GB<\/td>\n60.1GB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nMemory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model’s authors. Compare with other models in our VRAM quick reference<\/a>. What the quantization names mean: glossary<\/a>.<\/em><\/p>\n<\/p>\n<\/p>\nCan You Run It Locally?<\/h2>\nRuns in Ollama, LM Studio and llama.cpp as-is.<\/strong><\/p>\nIt is distributed in GGUF, so no conversion is needed.<\/p>\nLicense \u2014 cc-by-nc-4.0<\/code> (Commercial use prohibited):<\/strong> Commercial use is prohibited<\/strong> (NC = NonCommercial). Internal business use can also count as commercial, so avoid this license for work use.<\/p>\nCompression:<\/strong> the Q4_K_M build measures 5.56 bits per weight \u2014 about 35% the size of the original 16-bit weights, calculated by this site from the actual file sizes.<\/p>\nCompiled by this site’s code from the published formats, converted builds we have found, and each engine’s own model registry. “Not found” means we have not seen such a build, not that none exists. License summaries are not legal advice \u2014 check the publisher’s original terms before relying on them.<\/em><\/p>\n<\/p>\n<\/p>\nOur Own Measurements<\/h2>\nMeasurements We Did Not Take<\/h3>\nThis model’s license (cc-by-nc-4.0<\/code>) does not permit commercial use. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<\/p>\n<\/p>\nRecent Models in the Same Size Class<\/h2>\nModels with <\/em>15\u201340B<\/em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site’s estimates; licenses are as stated on the model cards.<\/em><\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nModel<\/th>\nParameters<\/th>\nSmallest VRAM tier<\/th>\nLicense<\/th>\nOur article<\/th>\n<\/tr>\n<\/thead>\n\n\nQwen\/Qwen3.8-27B<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Vision-Language Model: 80GB+ VRAM<\/a> (2026-10-02)<\/td>\n<\/tr>\n\nCloudflare\/clef<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Structured Decision-Making Model: 80GB+ VRAM<\/a> (2026-10-01)<\/td>\n<\/tr>\n\nQwen\/Qwen3.8-27B<\/td>\n27.8B<\/td>\n8GB<\/td>\napache-2.0<\/td>\nQwen3.8-27B Vision-Language Model: Our Test Answers, 8GB+ VRAM<\/a> (2026-09-26)<\/td>\n<\/tr>\n\nbartowski\/vectionlabs_Salience-27B-R6-GGUF<\/td>\n27.8B<\/td>\n12GB<\/td>\napache-2.0<\/td>\nvectionlabs_Salience-27B-R6-GGUF: Our Test Answers, 12GB+ VRAM<\/a> (2026-09-15)<\/td>\n<\/tr>\n\nagentionai\/Signal-3.8-27B-GGUF<\/td>\n27.8B<\/td>\n16GB<\/td>\napache-2.0<\/td>\nSignal-3.8-27B-GGUF: Our Test Answers, 16GB+ VRAM<\/a> (2026-09-12)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/p>\nHow to Get It<\/h2>\nThis model is distributed in GGUF format and can be run using inference engines such as llama.cpp<\/code>.<\/p>\nWhen launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
First, here is a comparison with other models across 10,000 text-based questions (using 34 public sources):<\/p>\n
From this table, we can see that OpenJev achieves an accuracy of 84.0%, significantly outperforming the pre-tuning base model (80.4%) and the open model Nimble 9B (75.7%). Meanwhile, while it fell slightly short of the hosted API Jev (hosted API) at 85.4%, it comes very close with a small difference of 1.4 points.<\/p>\n
Next is the breakdown of accuracy by category:<\/p>\n
\u2192 Scroll horizontally to see all columns<\/p>\n
Looking at these category results, OpenJev shows high accuracy in tasks such as “intent \/ routing \/ topic” (92.8%) and “reading + language” (89.2%). It also scores higher than the hosted API Jev in “ethics \/ policy judgement” (78.1%) and “sentiment \/ stance” (82.1%). However, in fields like “science \/ facts \/ claims” (82.5%) and “commonsense reasoning” (85.8%), it falls slightly behind the hosted API Jev (89.0% and 88.3% respectively).<\/p>\n
Additionally, the comparison results before and after tuning on agent-based decision-making tasks are as follows:<\/p>\n
From these results, OpenJev achieves a significant accuracy improvement of over 10 points compared to pre-tuning when determining next actions for desktop operations from screenshots or on unseen websites and domains. Furthermore, answer flips when options are shuffled decreased drastically from 18.5% to 2.3%, demonstrating extremely high stability as a decision-making model.<\/p>\n
Comparisons regarding multilingual support and long-context reading comprehension are as follows:<\/p>\n
This table indicates that accuracy has improved over pre-tuning in multilingual tasks involving German, French, Hindi, Chinese, and Japanese. On the other hand, in question tasks regarding long articles (QuALITY), the score remains unchanged at 91.7% before and after tuning, maintaining equivalent performance.<\/p>\n
Finally, here is the number of completed tasks when executing end-to-end browser tasks (100 types of MiniWoB tasks, 1 trial each):<\/p>\n
These results confirm that in end-to-end browser tasks, OpenJev completes 39 tasks, matching the hosted Jev API and showing a slight improvement from the 38 tasks completed before tuning.<\/p>\n
This model is a “decision model” that accepts text or images (screenshots) as inputs and makes optimal decisions from pre-defined choices. Unlike general chat bots that generate free text before parsing, it calculates probabilities directly from token scores at the initial output position, enabling fast and reliable processing.<\/p>\n
Based on the performance and features of the original model openjev\/openjev<\/code>, it is highly suited for the following applications and tasks:<\/p>\n\n\nRouting and Triage<\/strong>:\n Determining the intent, topic, responsible team, priority, or language of user inquiry messages to automatically route them to the appropriate destination.<\/p>\n<\/li>\n\nModeration and Safety Judgment<\/strong>:\n Evaluating whether input text falls under harmful content, spam, hate speech, policy violations, or legal threats.<\/p>\n<\/li>\n\nLLM as a Judge<\/strong>:\n Objectively determining whether a generative AI model’s output is based on reliable evidence (presence of hallucinations), follows provided evaluation criteria (rubrics), or which of multiple answers is superior.<\/p>\n<\/li>\n\nBusiness Processes and Document Management<\/strong>:\n Automating routine judgment tasks such as approving, withholding, or rejecting invoices, determining the severity of system alerts, or classifying specific clause types in contracts.<\/p>\n<\/li>\n\nBrowser and Desktop Agents<\/strong>:\n Reading screen screenshots, HTML (DOM structure), or JSON data to determine the next element to interact with or action to execute (such as clicks or input). It can also judge whether a task is complete or blocked.<\/p>\n<\/li>\n\nScoring<\/strong>:\n Calculating ordered step-by-step evaluations accompanied by expected values or confidence levels.<\/p>\n<\/li>\n<\/ul>\nFurthermore, because this model allows dynamic specification of choices (labels) per request, task-specific pre-training or fixed labeling is unnecessary. It can process up to 52 choices in a single request during a single forward pass, and can simultaneously handle multiple questions (e.g., routing, sentiment analysis, urgency) in one request. A major strength is its support for multiple languages including Japanese (English, German, French, Hindi, Chinese, and Japanese).<\/p>\nHow It Differs from Similar Models<\/h2>\nIn a previous article introduced on our site, Decision-Making Specialized Model “OpenJev” Testing GGUF Model Released<\/a>, we covered “tinyopenjev-for-testing-gguf”, also released by ggml-org. However, that was an ultra-small test-only model (approx. 34M parameters) created solely to verify whether inference engines and loaders function properly, and its output content carried no meaning.<\/p>\nIn contrast, the newly released “OpenJev-GGUF” is a practical quantized model that retains the entire structure of the original model openjev\/openjev<\/code>. Unlike the test model, it can be used directly for actual decision-making tasks and agent inference purposes.<\/p>\n<\/p>\nHardware Requirements<\/h2>\nEstimated requirements (calculated by Local Model Watch)<\/strong> \u2014 27.4B parameters (taken from the base model openjev\/openjev)<\/p>\n\n\n\n\nYour VRAM<\/th>\nQuantization<\/th>\nFile size<\/th>\nEst. memory needed<\/th>\n<\/tr>\n<\/thead>\n\n\n24GB (RTX 4090 \/ 3090, etc.)<\/td>\nQ4_K_M<\/td>\n17.7GB<\/td>\n21.2GB<\/td>\n<\/tr>\n\n32GB (RTX 5090, etc.)<\/td>\nQ8_0<\/td>\n26.6GB<\/td>\n32.0GB<\/td>\n<\/tr>\n\n80GB class (A100 \/ H100)<\/td>\nBF16<\/td>\n50.1GB<\/td>\n60.1GB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nMemory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model’s authors. Compare with other models in our VRAM quick reference<\/a>. What the quantization names mean: glossary<\/a>.<\/em><\/p>\n<\/p>\n<\/p>\nCan You Run It Locally?<\/h2>\nRuns in Ollama, LM Studio and llama.cpp as-is.<\/strong><\/p>\nIt is distributed in GGUF, so no conversion is needed.<\/p>\nLicense \u2014 cc-by-nc-4.0<\/code> (Commercial use prohibited):<\/strong> Commercial use is prohibited<\/strong> (NC = NonCommercial). Internal business use can also count as commercial, so avoid this license for work use.<\/p>\nCompression:<\/strong> the Q4_K_M build measures 5.56 bits per weight \u2014 about 35% the size of the original 16-bit weights, calculated by this site from the actual file sizes.<\/p>\nCompiled by this site’s code from the published formats, converted builds we have found, and each engine’s own model registry. “Not found” means we have not seen such a build, not that none exists. License summaries are not legal advice \u2014 check the publisher’s original terms before relying on them.<\/em><\/p>\n<\/p>\n<\/p>\nOur Own Measurements<\/h2>\nMeasurements We Did Not Take<\/h3>\nThis model’s license (cc-by-nc-4.0<\/code>) does not permit commercial use. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<\/p>\n<\/p>\nRecent Models in the Same Size Class<\/h2>\nModels with <\/em>15\u201340B<\/em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site’s estimates; licenses are as stated on the model cards.<\/em><\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nModel<\/th>\nParameters<\/th>\nSmallest VRAM tier<\/th>\nLicense<\/th>\nOur article<\/th>\n<\/tr>\n<\/thead>\n\n\nQwen\/Qwen3.8-27B<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Vision-Language Model: 80GB+ VRAM<\/a> (2026-10-02)<\/td>\n<\/tr>\n\nCloudflare\/clef<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Structured Decision-Making Model: 80GB+ VRAM<\/a> (2026-10-01)<\/td>\n<\/tr>\n\nQwen\/Qwen3.8-27B<\/td>\n27.8B<\/td>\n8GB<\/td>\napache-2.0<\/td>\nQwen3.8-27B Vision-Language Model: Our Test Answers, 8GB+ VRAM<\/a> (2026-09-26)<\/td>\n<\/tr>\n\nbartowski\/vectionlabs_Salience-27B-R6-GGUF<\/td>\n27.8B<\/td>\n12GB<\/td>\napache-2.0<\/td>\nvectionlabs_Salience-27B-R6-GGUF: Our Test Answers, 12GB+ VRAM<\/a> (2026-09-15)<\/td>\n<\/tr>\n\nagentionai\/Signal-3.8-27B-GGUF<\/td>\n27.8B<\/td>\n16GB<\/td>\napache-2.0<\/td>\nSignal-3.8-27B-GGUF: Our Test Answers, 16GB+ VRAM<\/a> (2026-09-12)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/p>\nHow to Get It<\/h2>\nThis model is distributed in GGUF format and can be run using inference engines such as llama.cpp<\/code>.<\/p>\nWhen launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
openjev\/openjev<\/code>, it is highly suited for the following applications and tasks:<\/p>\n\n\nRouting and Triage<\/strong>:\n Determining the intent, topic, responsible team, priority, or language of user inquiry messages to automatically route them to the appropriate destination.<\/p>\n<\/li>\n\nModeration and Safety Judgment<\/strong>:\n Evaluating whether input text falls under harmful content, spam, hate speech, policy violations, or legal threats.<\/p>\n<\/li>\n\nLLM as a Judge<\/strong>:\n Objectively determining whether a generative AI model’s output is based on reliable evidence (presence of hallucinations), follows provided evaluation criteria (rubrics), or which of multiple answers is superior.<\/p>\n<\/li>\n\nBusiness Processes and Document Management<\/strong>:\n Automating routine judgment tasks such as approving, withholding, or rejecting invoices, determining the severity of system alerts, or classifying specific clause types in contracts.<\/p>\n<\/li>\n\nBrowser and Desktop Agents<\/strong>:\n Reading screen screenshots, HTML (DOM structure), or JSON data to determine the next element to interact with or action to execute (such as clicks or input). It can also judge whether a task is complete or blocked.<\/p>\n<\/li>\n\nScoring<\/strong>:\n Calculating ordered step-by-step evaluations accompanied by expected values or confidence levels.<\/p>\n<\/li>\n<\/ul>\nFurthermore, because this model allows dynamic specification of choices (labels) per request, task-specific pre-training or fixed labeling is unnecessary. It can process up to 52 choices in a single request during a single forward pass, and can simultaneously handle multiple questions (e.g., routing, sentiment analysis, urgency) in one request. A major strength is its support for multiple languages including Japanese (English, German, French, Hindi, Chinese, and Japanese).<\/p>\nHow It Differs from Similar Models<\/h2>\nIn a previous article introduced on our site, Decision-Making Specialized Model “OpenJev” Testing GGUF Model Released<\/a>, we covered “tinyopenjev-for-testing-gguf”, also released by ggml-org. However, that was an ultra-small test-only model (approx. 34M parameters) created solely to verify whether inference engines and loaders function properly, and its output content carried no meaning.<\/p>\nIn contrast, the newly released “OpenJev-GGUF” is a practical quantized model that retains the entire structure of the original model openjev\/openjev<\/code>. Unlike the test model, it can be used directly for actual decision-making tasks and agent inference purposes.<\/p>\n<\/p>\nHardware Requirements<\/h2>\nEstimated requirements (calculated by Local Model Watch)<\/strong> \u2014 27.4B parameters (taken from the base model openjev\/openjev)<\/p>\n\n\n\n\nYour VRAM<\/th>\nQuantization<\/th>\nFile size<\/th>\nEst. memory needed<\/th>\n<\/tr>\n<\/thead>\n\n\n24GB (RTX 4090 \/ 3090, etc.)<\/td>\nQ4_K_M<\/td>\n17.7GB<\/td>\n21.2GB<\/td>\n<\/tr>\n\n32GB (RTX 5090, etc.)<\/td>\nQ8_0<\/td>\n26.6GB<\/td>\n32.0GB<\/td>\n<\/tr>\n\n80GB class (A100 \/ H100)<\/td>\nBF16<\/td>\n50.1GB<\/td>\n60.1GB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nMemory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model’s authors. Compare with other models in our VRAM quick reference<\/a>. What the quantization names mean: glossary<\/a>.<\/em><\/p>\n<\/p>\n<\/p>\nCan You Run It Locally?<\/h2>\nRuns in Ollama, LM Studio and llama.cpp as-is.<\/strong><\/p>\nIt is distributed in GGUF, so no conversion is needed.<\/p>\nLicense \u2014 cc-by-nc-4.0<\/code> (Commercial use prohibited):<\/strong> Commercial use is prohibited<\/strong> (NC = NonCommercial). Internal business use can also count as commercial, so avoid this license for work use.<\/p>\nCompression:<\/strong> the Q4_K_M build measures 5.56 bits per weight \u2014 about 35% the size of the original 16-bit weights, calculated by this site from the actual file sizes.<\/p>\nCompiled by this site’s code from the published formats, converted builds we have found, and each engine’s own model registry. “Not found” means we have not seen such a build, not that none exists. License summaries are not legal advice \u2014 check the publisher’s original terms before relying on them.<\/em><\/p>\n<\/p>\n<\/p>\nOur Own Measurements<\/h2>\nMeasurements We Did Not Take<\/h3>\nThis model’s license (cc-by-nc-4.0<\/code>) does not permit commercial use. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<\/p>\n<\/p>\nRecent Models in the Same Size Class<\/h2>\nModels with <\/em>15\u201340B<\/em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site’s estimates; licenses are as stated on the model cards.<\/em><\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nModel<\/th>\nParameters<\/th>\nSmallest VRAM tier<\/th>\nLicense<\/th>\nOur article<\/th>\n<\/tr>\n<\/thead>\n\n\nQwen\/Qwen3.8-27B<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Vision-Language Model: 80GB+ VRAM<\/a> (2026-10-02)<\/td>\n<\/tr>\n\nCloudflare\/clef<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Structured Decision-Making Model: 80GB+ VRAM<\/a> (2026-10-01)<\/td>\n<\/tr>\n\nQwen\/Qwen3.8-27B<\/td>\n27.8B<\/td>\n8GB<\/td>\napache-2.0<\/td>\nQwen3.8-27B Vision-Language Model: Our Test Answers, 8GB+ VRAM<\/a> (2026-09-26)<\/td>\n<\/tr>\n\nbartowski\/vectionlabs_Salience-27B-R6-GGUF<\/td>\n27.8B<\/td>\n12GB<\/td>\napache-2.0<\/td>\nvectionlabs_Salience-27B-R6-GGUF: Our Test Answers, 12GB+ VRAM<\/a> (2026-09-15)<\/td>\n<\/tr>\n\nagentionai\/Signal-3.8-27B-GGUF<\/td>\n27.8B<\/td>\n16GB<\/td>\napache-2.0<\/td>\nSignal-3.8-27B-GGUF: Our Test Answers, 16GB+ VRAM<\/a> (2026-09-12)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/p>\nHow to Get It<\/h2>\nThis model is distributed in GGUF format and can be run using inference engines such as llama.cpp<\/code>.<\/p>\nWhen launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
Routing and Triage<\/strong>:\n Determining the intent, topic, responsible team, priority, or language of user inquiry messages to automatically route them to the appropriate destination.<\/p>\n<\/li>\n\nModeration and Safety Judgment<\/strong>:\n Evaluating whether input text falls under harmful content, spam, hate speech, policy violations, or legal threats.<\/p>\n<\/li>\n\nLLM as a Judge<\/strong>:\n Objectively determining whether a generative AI model’s output is based on reliable evidence (presence of hallucinations), follows provided evaluation criteria (rubrics), or which of multiple answers is superior.<\/p>\n<\/li>\n\nBusiness Processes and Document Management<\/strong>:\n Automating routine judgment tasks such as approving, withholding, or rejecting invoices, determining the severity of system alerts, or classifying specific clause types in contracts.<\/p>\n<\/li>\n\nBrowser and Desktop Agents<\/strong>:\n Reading screen screenshots, HTML (DOM structure), or JSON data to determine the next element to interact with or action to execute (such as clicks or input). It can also judge whether a task is complete or blocked.<\/p>\n<\/li>\n\nScoring<\/strong>:\n Calculating ordered step-by-step evaluations accompanied by expected values or confidence levels.<\/p>\n<\/li>\n<\/ul>\nFurthermore, because this model allows dynamic specification of choices (labels) per request, task-specific pre-training or fixed labeling is unnecessary. It can process up to 52 choices in a single request during a single forward pass, and can simultaneously handle multiple questions (e.g., routing, sentiment analysis, urgency) in one request. A major strength is its support for multiple languages including Japanese (English, German, French, Hindi, Chinese, and Japanese).<\/p>\nHow It Differs from Similar Models<\/h2>\nIn a previous article introduced on our site, Decision-Making Specialized Model “OpenJev” Testing GGUF Model Released<\/a>, we covered “tinyopenjev-for-testing-gguf”, also released by ggml-org. However, that was an ultra-small test-only model (approx. 34M parameters) created solely to verify whether inference engines and loaders function properly, and its output content carried no meaning.<\/p>\nIn contrast, the newly released “OpenJev-GGUF” is a practical quantized model that retains the entire structure of the original model openjev\/openjev<\/code>. Unlike the test model, it can be used directly for actual decision-making tasks and agent inference purposes.<\/p>\n<\/p>\nHardware Requirements<\/h2>\nEstimated requirements (calculated by Local Model Watch)<\/strong> \u2014 27.4B parameters (taken from the base model openjev\/openjev)<\/p>\n\n\n\n\nYour VRAM<\/th>\nQuantization<\/th>\nFile size<\/th>\nEst. memory needed<\/th>\n<\/tr>\n<\/thead>\n\n\n24GB (RTX 4090 \/ 3090, etc.)<\/td>\nQ4_K_M<\/td>\n17.7GB<\/td>\n21.2GB<\/td>\n<\/tr>\n\n32GB (RTX 5090, etc.)<\/td>\nQ8_0<\/td>\n26.6GB<\/td>\n32.0GB<\/td>\n<\/tr>\n\n80GB class (A100 \/ H100)<\/td>\nBF16<\/td>\n50.1GB<\/td>\n60.1GB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nMemory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model’s authors. Compare with other models in our VRAM quick reference<\/a>. What the quantization names mean: glossary<\/a>.<\/em><\/p>\n<\/p>\n<\/p>\nCan You Run It Locally?<\/h2>\nRuns in Ollama, LM Studio and llama.cpp as-is.<\/strong><\/p>\nIt is distributed in GGUF, so no conversion is needed.<\/p>\nLicense \u2014 cc-by-nc-4.0<\/code> (Commercial use prohibited):<\/strong> Commercial use is prohibited<\/strong> (NC = NonCommercial). Internal business use can also count as commercial, so avoid this license for work use.<\/p>\nCompression:<\/strong> the Q4_K_M build measures 5.56 bits per weight \u2014 about 35% the size of the original 16-bit weights, calculated by this site from the actual file sizes.<\/p>\nCompiled by this site’s code from the published formats, converted builds we have found, and each engine’s own model registry. “Not found” means we have not seen such a build, not that none exists. License summaries are not legal advice \u2014 check the publisher’s original terms before relying on them.<\/em><\/p>\n<\/p>\n<\/p>\nOur Own Measurements<\/h2>\nMeasurements We Did Not Take<\/h3>\nThis model’s license (cc-by-nc-4.0<\/code>) does not permit commercial use. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<\/p>\n<\/p>\nRecent Models in the Same Size Class<\/h2>\nModels with <\/em>15\u201340B<\/em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site’s estimates; licenses are as stated on the model cards.<\/em><\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nModel<\/th>\nParameters<\/th>\nSmallest VRAM tier<\/th>\nLicense<\/th>\nOur article<\/th>\n<\/tr>\n<\/thead>\n\n\nQwen\/Qwen3.8-27B<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Vision-Language Model: 80GB+ VRAM<\/a> (2026-10-02)<\/td>\n<\/tr>\n\nCloudflare\/clef<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Structured Decision-Making Model: 80GB+ VRAM<\/a> (2026-10-01)<\/td>\n<\/tr>\n\nQwen\/Qwen3.8-27B<\/td>\n27.8B<\/td>\n8GB<\/td>\napache-2.0<\/td>\nQwen3.8-27B Vision-Language Model: Our Test Answers, 8GB+ VRAM<\/a> (2026-09-26)<\/td>\n<\/tr>\n\nbartowski\/vectionlabs_Salience-27B-R6-GGUF<\/td>\n27.8B<\/td>\n12GB<\/td>\napache-2.0<\/td>\nvectionlabs_Salience-27B-R6-GGUF: Our Test Answers, 12GB+ VRAM<\/a> (2026-09-15)<\/td>\n<\/tr>\n\nagentionai\/Signal-3.8-27B-GGUF<\/td>\n27.8B<\/td>\n16GB<\/td>\napache-2.0<\/td>\nSignal-3.8-27B-GGUF: Our Test Answers, 16GB+ VRAM<\/a> (2026-09-12)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/p>\nHow to Get It<\/h2>\nThis model is distributed in GGUF format and can be run using inference engines such as llama.cpp<\/code>.<\/p>\nWhen launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
Moderation and Safety Judgment<\/strong>:\n Evaluating whether input text falls under harmful content, spam, hate speech, policy violations, or legal threats.<\/p>\n<\/li>\n\nLLM as a Judge<\/strong>:\n Objectively determining whether a generative AI model’s output is based on reliable evidence (presence of hallucinations), follows provided evaluation criteria (rubrics), or which of multiple answers is superior.<\/p>\n<\/li>\n\nBusiness Processes and Document Management<\/strong>:\n Automating routine judgment tasks such as approving, withholding, or rejecting invoices, determining the severity of system alerts, or classifying specific clause types in contracts.<\/p>\n<\/li>\n\nBrowser and Desktop Agents<\/strong>:\n Reading screen screenshots, HTML (DOM structure), or JSON data to determine the next element to interact with or action to execute (such as clicks or input). It can also judge whether a task is complete or blocked.<\/p>\n<\/li>\n\nScoring<\/strong>:\n Calculating ordered step-by-step evaluations accompanied by expected values or confidence levels.<\/p>\n<\/li>\n<\/ul>\nFurthermore, because this model allows dynamic specification of choices (labels) per request, task-specific pre-training or fixed labeling is unnecessary. It can process up to 52 choices in a single request during a single forward pass, and can simultaneously handle multiple questions (e.g., routing, sentiment analysis, urgency) in one request. A major strength is its support for multiple languages including Japanese (English, German, French, Hindi, Chinese, and Japanese).<\/p>\nHow It Differs from Similar Models<\/h2>\nIn a previous article introduced on our site, Decision-Making Specialized Model “OpenJev” Testing GGUF Model Released<\/a>, we covered “tinyopenjev-for-testing-gguf”, also released by ggml-org. However, that was an ultra-small test-only model (approx. 34M parameters) created solely to verify whether inference engines and loaders function properly, and its output content carried no meaning.<\/p>\nIn contrast, the newly released “OpenJev-GGUF” is a practical quantized model that retains the entire structure of the original model openjev\/openjev<\/code>. Unlike the test model, it can be used directly for actual decision-making tasks and agent inference purposes.<\/p>\n<\/p>\nHardware Requirements<\/h2>\nEstimated requirements (calculated by Local Model Watch)<\/strong> \u2014 27.4B parameters (taken from the base model openjev\/openjev)<\/p>\n\n\n\n\nYour VRAM<\/th>\nQuantization<\/th>\nFile size<\/th>\nEst. memory needed<\/th>\n<\/tr>\n<\/thead>\n\n\n24GB (RTX 4090 \/ 3090, etc.)<\/td>\nQ4_K_M<\/td>\n17.7GB<\/td>\n21.2GB<\/td>\n<\/tr>\n\n32GB (RTX 5090, etc.)<\/td>\nQ8_0<\/td>\n26.6GB<\/td>\n32.0GB<\/td>\n<\/tr>\n\n80GB class (A100 \/ H100)<\/td>\nBF16<\/td>\n50.1GB<\/td>\n60.1GB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nMemory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model’s authors. Compare with other models in our VRAM quick reference<\/a>. What the quantization names mean: glossary<\/a>.<\/em><\/p>\n<\/p>\n<\/p>\nCan You Run It Locally?<\/h2>\nRuns in Ollama, LM Studio and llama.cpp as-is.<\/strong><\/p>\nIt is distributed in GGUF, so no conversion is needed.<\/p>\nLicense \u2014 cc-by-nc-4.0<\/code> (Commercial use prohibited):<\/strong> Commercial use is prohibited<\/strong> (NC = NonCommercial). Internal business use can also count as commercial, so avoid this license for work use.<\/p>\nCompression:<\/strong> the Q4_K_M build measures 5.56 bits per weight \u2014 about 35% the size of the original 16-bit weights, calculated by this site from the actual file sizes.<\/p>\nCompiled by this site’s code from the published formats, converted builds we have found, and each engine’s own model registry. “Not found” means we have not seen such a build, not that none exists. License summaries are not legal advice \u2014 check the publisher’s original terms before relying on them.<\/em><\/p>\n<\/p>\n<\/p>\nOur Own Measurements<\/h2>\nMeasurements We Did Not Take<\/h3>\nThis model’s license (cc-by-nc-4.0<\/code>) does not permit commercial use. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<\/p>\n<\/p>\nRecent Models in the Same Size Class<\/h2>\nModels with <\/em>15\u201340B<\/em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site’s estimates; licenses are as stated on the model cards.<\/em><\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nModel<\/th>\nParameters<\/th>\nSmallest VRAM tier<\/th>\nLicense<\/th>\nOur article<\/th>\n<\/tr>\n<\/thead>\n\n\nQwen\/Qwen3.8-27B<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Vision-Language Model: 80GB+ VRAM<\/a> (2026-10-02)<\/td>\n<\/tr>\n\nCloudflare\/clef<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Structured Decision-Making Model: 80GB+ VRAM<\/a> (2026-10-01)<\/td>\n<\/tr>\n\nQwen\/Qwen3.8-27B<\/td>\n27.8B<\/td>\n8GB<\/td>\napache-2.0<\/td>\nQwen3.8-27B Vision-Language Model: Our Test Answers, 8GB+ VRAM<\/a> (2026-09-26)<\/td>\n<\/tr>\n\nbartowski\/vectionlabs_Salience-27B-R6-GGUF<\/td>\n27.8B<\/td>\n12GB<\/td>\napache-2.0<\/td>\nvectionlabs_Salience-27B-R6-GGUF: Our Test Answers, 12GB+ VRAM<\/a> (2026-09-15)<\/td>\n<\/tr>\n\nagentionai\/Signal-3.8-27B-GGUF<\/td>\n27.8B<\/td>\n16GB<\/td>\napache-2.0<\/td>\nSignal-3.8-27B-GGUF: Our Test Answers, 16GB+ VRAM<\/a> (2026-09-12)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/p>\nHow to Get It<\/h2>\nThis model is distributed in GGUF format and can be run using inference engines such as llama.cpp<\/code>.<\/p>\nWhen launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
LLM as a Judge<\/strong>:\n Objectively determining whether a generative AI model’s output is based on reliable evidence (presence of hallucinations), follows provided evaluation criteria (rubrics), or which of multiple answers is superior.<\/p>\n<\/li>\n\nBusiness Processes and Document Management<\/strong>:\n Automating routine judgment tasks such as approving, withholding, or rejecting invoices, determining the severity of system alerts, or classifying specific clause types in contracts.<\/p>\n<\/li>\n\nBrowser and Desktop Agents<\/strong>:\n Reading screen screenshots, HTML (DOM structure), or JSON data to determine the next element to interact with or action to execute (such as clicks or input). It can also judge whether a task is complete or blocked.<\/p>\n<\/li>\n\nScoring<\/strong>:\n Calculating ordered step-by-step evaluations accompanied by expected values or confidence levels.<\/p>\n<\/li>\n<\/ul>\nFurthermore, because this model allows dynamic specification of choices (labels) per request, task-specific pre-training or fixed labeling is unnecessary. It can process up to 52 choices in a single request during a single forward pass, and can simultaneously handle multiple questions (e.g., routing, sentiment analysis, urgency) in one request. A major strength is its support for multiple languages including Japanese (English, German, French, Hindi, Chinese, and Japanese).<\/p>\nHow It Differs from Similar Models<\/h2>\nIn a previous article introduced on our site, Decision-Making Specialized Model “OpenJev” Testing GGUF Model Released<\/a>, we covered “tinyopenjev-for-testing-gguf”, also released by ggml-org. However, that was an ultra-small test-only model (approx. 34M parameters) created solely to verify whether inference engines and loaders function properly, and its output content carried no meaning.<\/p>\nIn contrast, the newly released “OpenJev-GGUF” is a practical quantized model that retains the entire structure of the original model openjev\/openjev<\/code>. Unlike the test model, it can be used directly for actual decision-making tasks and agent inference purposes.<\/p>\n<\/p>\nHardware Requirements<\/h2>\nEstimated requirements (calculated by Local Model Watch)<\/strong> \u2014 27.4B parameters (taken from the base model openjev\/openjev)<\/p>\n\n\n\n\nYour VRAM<\/th>\nQuantization<\/th>\nFile size<\/th>\nEst. memory needed<\/th>\n<\/tr>\n<\/thead>\n\n\n24GB (RTX 4090 \/ 3090, etc.)<\/td>\nQ4_K_M<\/td>\n17.7GB<\/td>\n21.2GB<\/td>\n<\/tr>\n\n32GB (RTX 5090, etc.)<\/td>\nQ8_0<\/td>\n26.6GB<\/td>\n32.0GB<\/td>\n<\/tr>\n\n80GB class (A100 \/ H100)<\/td>\nBF16<\/td>\n50.1GB<\/td>\n60.1GB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nMemory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model’s authors. Compare with other models in our VRAM quick reference<\/a>. What the quantization names mean: glossary<\/a>.<\/em><\/p>\n<\/p>\n<\/p>\nCan You Run It Locally?<\/h2>\nRuns in Ollama, LM Studio and llama.cpp as-is.<\/strong><\/p>\nIt is distributed in GGUF, so no conversion is needed.<\/p>\nLicense \u2014 cc-by-nc-4.0<\/code> (Commercial use prohibited):<\/strong> Commercial use is prohibited<\/strong> (NC = NonCommercial). Internal business use can also count as commercial, so avoid this license for work use.<\/p>\nCompression:<\/strong> the Q4_K_M build measures 5.56 bits per weight \u2014 about 35% the size of the original 16-bit weights, calculated by this site from the actual file sizes.<\/p>\nCompiled by this site’s code from the published formats, converted builds we have found, and each engine’s own model registry. “Not found” means we have not seen such a build, not that none exists. License summaries are not legal advice \u2014 check the publisher’s original terms before relying on them.<\/em><\/p>\n<\/p>\n<\/p>\nOur Own Measurements<\/h2>\nMeasurements We Did Not Take<\/h3>\nThis model’s license (cc-by-nc-4.0<\/code>) does not permit commercial use. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<\/p>\n<\/p>\nRecent Models in the Same Size Class<\/h2>\nModels with <\/em>15\u201340B<\/em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site’s estimates; licenses are as stated on the model cards.<\/em><\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nModel<\/th>\nParameters<\/th>\nSmallest VRAM tier<\/th>\nLicense<\/th>\nOur article<\/th>\n<\/tr>\n<\/thead>\n\n\nQwen\/Qwen3.8-27B<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Vision-Language Model: 80GB+ VRAM<\/a> (2026-10-02)<\/td>\n<\/tr>\n\nCloudflare\/clef<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Structured Decision-Making Model: 80GB+ VRAM<\/a> (2026-10-01)<\/td>\n<\/tr>\n\nQwen\/Qwen3.8-27B<\/td>\n27.8B<\/td>\n8GB<\/td>\napache-2.0<\/td>\nQwen3.8-27B Vision-Language Model: Our Test Answers, 8GB+ VRAM<\/a> (2026-09-26)<\/td>\n<\/tr>\n\nbartowski\/vectionlabs_Salience-27B-R6-GGUF<\/td>\n27.8B<\/td>\n12GB<\/td>\napache-2.0<\/td>\nvectionlabs_Salience-27B-R6-GGUF: Our Test Answers, 12GB+ VRAM<\/a> (2026-09-15)<\/td>\n<\/tr>\n\nagentionai\/Signal-3.8-27B-GGUF<\/td>\n27.8B<\/td>\n16GB<\/td>\napache-2.0<\/td>\nSignal-3.8-27B-GGUF: Our Test Answers, 16GB+ VRAM<\/a> (2026-09-12)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/p>\nHow to Get It<\/h2>\nThis model is distributed in GGUF format and can be run using inference engines such as llama.cpp<\/code>.<\/p>\nWhen launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
Business Processes and Document Management<\/strong>:\n Automating routine judgment tasks such as approving, withholding, or rejecting invoices, determining the severity of system alerts, or classifying specific clause types in contracts.<\/p>\n<\/li>\n\nBrowser and Desktop Agents<\/strong>:\n Reading screen screenshots, HTML (DOM structure), or JSON data to determine the next element to interact with or action to execute (such as clicks or input). It can also judge whether a task is complete or blocked.<\/p>\n<\/li>\n\nScoring<\/strong>:\n Calculating ordered step-by-step evaluations accompanied by expected values or confidence levels.<\/p>\n<\/li>\n<\/ul>\nFurthermore, because this model allows dynamic specification of choices (labels) per request, task-specific pre-training or fixed labeling is unnecessary. It can process up to 52 choices in a single request during a single forward pass, and can simultaneously handle multiple questions (e.g., routing, sentiment analysis, urgency) in one request. A major strength is its support for multiple languages including Japanese (English, German, French, Hindi, Chinese, and Japanese).<\/p>\nHow It Differs from Similar Models<\/h2>\nIn a previous article introduced on our site, Decision-Making Specialized Model “OpenJev” Testing GGUF Model Released<\/a>, we covered “tinyopenjev-for-testing-gguf”, also released by ggml-org. However, that was an ultra-small test-only model (approx. 34M parameters) created solely to verify whether inference engines and loaders function properly, and its output content carried no meaning.<\/p>\nIn contrast, the newly released “OpenJev-GGUF” is a practical quantized model that retains the entire structure of the original model openjev\/openjev<\/code>. Unlike the test model, it can be used directly for actual decision-making tasks and agent inference purposes.<\/p>\n<\/p>\nHardware Requirements<\/h2>\nEstimated requirements (calculated by Local Model Watch)<\/strong> \u2014 27.4B parameters (taken from the base model openjev\/openjev)<\/p>\n\n\n\n\nYour VRAM<\/th>\nQuantization<\/th>\nFile size<\/th>\nEst. memory needed<\/th>\n<\/tr>\n<\/thead>\n\n\n24GB (RTX 4090 \/ 3090, etc.)<\/td>\nQ4_K_M<\/td>\n17.7GB<\/td>\n21.2GB<\/td>\n<\/tr>\n\n32GB (RTX 5090, etc.)<\/td>\nQ8_0<\/td>\n26.6GB<\/td>\n32.0GB<\/td>\n<\/tr>\n\n80GB class (A100 \/ H100)<\/td>\nBF16<\/td>\n50.1GB<\/td>\n60.1GB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nMemory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model’s authors. Compare with other models in our VRAM quick reference<\/a>. What the quantization names mean: glossary<\/a>.<\/em><\/p>\n<\/p>\n<\/p>\nCan You Run It Locally?<\/h2>\nRuns in Ollama, LM Studio and llama.cpp as-is.<\/strong><\/p>\nIt is distributed in GGUF, so no conversion is needed.<\/p>\nLicense \u2014 cc-by-nc-4.0<\/code> (Commercial use prohibited):<\/strong> Commercial use is prohibited<\/strong> (NC = NonCommercial). Internal business use can also count as commercial, so avoid this license for work use.<\/p>\nCompression:<\/strong> the Q4_K_M build measures 5.56 bits per weight \u2014 about 35% the size of the original 16-bit weights, calculated by this site from the actual file sizes.<\/p>\nCompiled by this site’s code from the published formats, converted builds we have found, and each engine’s own model registry. “Not found” means we have not seen such a build, not that none exists. License summaries are not legal advice \u2014 check the publisher’s original terms before relying on them.<\/em><\/p>\n<\/p>\n<\/p>\nOur Own Measurements<\/h2>\nMeasurements We Did Not Take<\/h3>\nThis model’s license (cc-by-nc-4.0<\/code>) does not permit commercial use. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<\/p>\n<\/p>\nRecent Models in the Same Size Class<\/h2>\nModels with <\/em>15\u201340B<\/em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site’s estimates; licenses are as stated on the model cards.<\/em><\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nModel<\/th>\nParameters<\/th>\nSmallest VRAM tier<\/th>\nLicense<\/th>\nOur article<\/th>\n<\/tr>\n<\/thead>\n\n\nQwen\/Qwen3.8-27B<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Vision-Language Model: 80GB+ VRAM<\/a> (2026-10-02)<\/td>\n<\/tr>\n\nCloudflare\/clef<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Structured Decision-Making Model: 80GB+ VRAM<\/a> (2026-10-01)<\/td>\n<\/tr>\n\nQwen\/Qwen3.8-27B<\/td>\n27.8B<\/td>\n8GB<\/td>\napache-2.0<\/td>\nQwen3.8-27B Vision-Language Model: Our Test Answers, 8GB+ VRAM<\/a> (2026-09-26)<\/td>\n<\/tr>\n\nbartowski\/vectionlabs_Salience-27B-R6-GGUF<\/td>\n27.8B<\/td>\n12GB<\/td>\napache-2.0<\/td>\nvectionlabs_Salience-27B-R6-GGUF: Our Test Answers, 12GB+ VRAM<\/a> (2026-09-15)<\/td>\n<\/tr>\n\nagentionai\/Signal-3.8-27B-GGUF<\/td>\n27.8B<\/td>\n16GB<\/td>\napache-2.0<\/td>\nSignal-3.8-27B-GGUF: Our Test Answers, 16GB+ VRAM<\/a> (2026-09-12)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/p>\nHow to Get It<\/h2>\nThis model is distributed in GGUF format and can be run using inference engines such as llama.cpp<\/code>.<\/p>\nWhen launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
Browser and Desktop Agents<\/strong>:\n Reading screen screenshots, HTML (DOM structure), or JSON data to determine the next element to interact with or action to execute (such as clicks or input). It can also judge whether a task is complete or blocked.<\/p>\n<\/li>\n\nScoring<\/strong>:\n Calculating ordered step-by-step evaluations accompanied by expected values or confidence levels.<\/p>\n<\/li>\n<\/ul>\nFurthermore, because this model allows dynamic specification of choices (labels) per request, task-specific pre-training or fixed labeling is unnecessary. It can process up to 52 choices in a single request during a single forward pass, and can simultaneously handle multiple questions (e.g., routing, sentiment analysis, urgency) in one request. A major strength is its support for multiple languages including Japanese (English, German, French, Hindi, Chinese, and Japanese).<\/p>\nHow It Differs from Similar Models<\/h2>\nIn a previous article introduced on our site, Decision-Making Specialized Model “OpenJev” Testing GGUF Model Released<\/a>, we covered “tinyopenjev-for-testing-gguf”, also released by ggml-org. However, that was an ultra-small test-only model (approx. 34M parameters) created solely to verify whether inference engines and loaders function properly, and its output content carried no meaning.<\/p>\nIn contrast, the newly released “OpenJev-GGUF” is a practical quantized model that retains the entire structure of the original model openjev\/openjev<\/code>. Unlike the test model, it can be used directly for actual decision-making tasks and agent inference purposes.<\/p>\n<\/p>\nHardware Requirements<\/h2>\nEstimated requirements (calculated by Local Model Watch)<\/strong> \u2014 27.4B parameters (taken from the base model openjev\/openjev)<\/p>\n\n\n\n\nYour VRAM<\/th>\nQuantization<\/th>\nFile size<\/th>\nEst. memory needed<\/th>\n<\/tr>\n<\/thead>\n\n\n24GB (RTX 4090 \/ 3090, etc.)<\/td>\nQ4_K_M<\/td>\n17.7GB<\/td>\n21.2GB<\/td>\n<\/tr>\n\n32GB (RTX 5090, etc.)<\/td>\nQ8_0<\/td>\n26.6GB<\/td>\n32.0GB<\/td>\n<\/tr>\n\n80GB class (A100 \/ H100)<\/td>\nBF16<\/td>\n50.1GB<\/td>\n60.1GB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nMemory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model’s authors. Compare with other models in our VRAM quick reference<\/a>. What the quantization names mean: glossary<\/a>.<\/em><\/p>\n<\/p>\n<\/p>\nCan You Run It Locally?<\/h2>\nRuns in Ollama, LM Studio and llama.cpp as-is.<\/strong><\/p>\nIt is distributed in GGUF, so no conversion is needed.<\/p>\nLicense \u2014 cc-by-nc-4.0<\/code> (Commercial use prohibited):<\/strong> Commercial use is prohibited<\/strong> (NC = NonCommercial). Internal business use can also count as commercial, so avoid this license for work use.<\/p>\nCompression:<\/strong> the Q4_K_M build measures 5.56 bits per weight \u2014 about 35% the size of the original 16-bit weights, calculated by this site from the actual file sizes.<\/p>\nCompiled by this site’s code from the published formats, converted builds we have found, and each engine’s own model registry. “Not found” means we have not seen such a build, not that none exists. License summaries are not legal advice \u2014 check the publisher’s original terms before relying on them.<\/em><\/p>\n<\/p>\n<\/p>\nOur Own Measurements<\/h2>\nMeasurements We Did Not Take<\/h3>\nThis model’s license (cc-by-nc-4.0<\/code>) does not permit commercial use. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<\/p>\n<\/p>\nRecent Models in the Same Size Class<\/h2>\nModels with <\/em>15\u201340B<\/em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site’s estimates; licenses are as stated on the model cards.<\/em><\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nModel<\/th>\nParameters<\/th>\nSmallest VRAM tier<\/th>\nLicense<\/th>\nOur article<\/th>\n<\/tr>\n<\/thead>\n\n\nQwen\/Qwen3.8-27B<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Vision-Language Model: 80GB+ VRAM<\/a> (2026-10-02)<\/td>\n<\/tr>\n\nCloudflare\/clef<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Structured Decision-Making Model: 80GB+ VRAM<\/a> (2026-10-01)<\/td>\n<\/tr>\n\nQwen\/Qwen3.8-27B<\/td>\n27.8B<\/td>\n8GB<\/td>\napache-2.0<\/td>\nQwen3.8-27B Vision-Language Model: Our Test Answers, 8GB+ VRAM<\/a> (2026-09-26)<\/td>\n<\/tr>\n\nbartowski\/vectionlabs_Salience-27B-R6-GGUF<\/td>\n27.8B<\/td>\n12GB<\/td>\napache-2.0<\/td>\nvectionlabs_Salience-27B-R6-GGUF: Our Test Answers, 12GB+ VRAM<\/a> (2026-09-15)<\/td>\n<\/tr>\n\nagentionai\/Signal-3.8-27B-GGUF<\/td>\n27.8B<\/td>\n16GB<\/td>\napache-2.0<\/td>\nSignal-3.8-27B-GGUF: Our Test Answers, 16GB+ VRAM<\/a> (2026-09-12)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/p>\nHow to Get It<\/h2>\nThis model is distributed in GGUF format and can be run using inference engines such as llama.cpp<\/code>.<\/p>\nWhen launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
Scoring<\/strong>:\n Calculating ordered step-by-step evaluations accompanied by expected values or confidence levels.<\/p>\n<\/li>\n<\/ul>\nFurthermore, because this model allows dynamic specification of choices (labels) per request, task-specific pre-training or fixed labeling is unnecessary. It can process up to 52 choices in a single request during a single forward pass, and can simultaneously handle multiple questions (e.g., routing, sentiment analysis, urgency) in one request. A major strength is its support for multiple languages including Japanese (English, German, French, Hindi, Chinese, and Japanese).<\/p>\nHow It Differs from Similar Models<\/h2>\nIn a previous article introduced on our site, Decision-Making Specialized Model “OpenJev” Testing GGUF Model Released<\/a>, we covered “tinyopenjev-for-testing-gguf”, also released by ggml-org. However, that was an ultra-small test-only model (approx. 34M parameters) created solely to verify whether inference engines and loaders function properly, and its output content carried no meaning.<\/p>\nIn contrast, the newly released “OpenJev-GGUF” is a practical quantized model that retains the entire structure of the original model openjev\/openjev<\/code>. Unlike the test model, it can be used directly for actual decision-making tasks and agent inference purposes.<\/p>\n<\/p>\nHardware Requirements<\/h2>\nEstimated requirements (calculated by Local Model Watch)<\/strong> \u2014 27.4B parameters (taken from the base model openjev\/openjev)<\/p>\n\n\n\n\nYour VRAM<\/th>\nQuantization<\/th>\nFile size<\/th>\nEst. memory needed<\/th>\n<\/tr>\n<\/thead>\n\n\n24GB (RTX 4090 \/ 3090, etc.)<\/td>\nQ4_K_M<\/td>\n17.7GB<\/td>\n21.2GB<\/td>\n<\/tr>\n\n32GB (RTX 5090, etc.)<\/td>\nQ8_0<\/td>\n26.6GB<\/td>\n32.0GB<\/td>\n<\/tr>\n\n80GB class (A100 \/ H100)<\/td>\nBF16<\/td>\n50.1GB<\/td>\n60.1GB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nMemory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model’s authors. Compare with other models in our VRAM quick reference<\/a>. What the quantization names mean: glossary<\/a>.<\/em><\/p>\n<\/p>\n<\/p>\nCan You Run It Locally?<\/h2>\nRuns in Ollama, LM Studio and llama.cpp as-is.<\/strong><\/p>\nIt is distributed in GGUF, so no conversion is needed.<\/p>\nLicense \u2014 cc-by-nc-4.0<\/code> (Commercial use prohibited):<\/strong> Commercial use is prohibited<\/strong> (NC = NonCommercial). Internal business use can also count as commercial, so avoid this license for work use.<\/p>\nCompression:<\/strong> the Q4_K_M build measures 5.56 bits per weight \u2014 about 35% the size of the original 16-bit weights, calculated by this site from the actual file sizes.<\/p>\nCompiled by this site’s code from the published formats, converted builds we have found, and each engine’s own model registry. “Not found” means we have not seen such a build, not that none exists. License summaries are not legal advice \u2014 check the publisher’s original terms before relying on them.<\/em><\/p>\n<\/p>\n<\/p>\nOur Own Measurements<\/h2>\nMeasurements We Did Not Take<\/h3>\nThis model’s license (cc-by-nc-4.0<\/code>) does not permit commercial use. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<\/p>\n<\/p>\nRecent Models in the Same Size Class<\/h2>\nModels with <\/em>15\u201340B<\/em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site’s estimates; licenses are as stated on the model cards.<\/em><\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nModel<\/th>\nParameters<\/th>\nSmallest VRAM tier<\/th>\nLicense<\/th>\nOur article<\/th>\n<\/tr>\n<\/thead>\n\n\nQwen\/Qwen3.8-27B<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Vision-Language Model: 80GB+ VRAM<\/a> (2026-10-02)<\/td>\n<\/tr>\n\nCloudflare\/clef<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Structured Decision-Making Model: 80GB+ VRAM<\/a> (2026-10-01)<\/td>\n<\/tr>\n\nQwen\/Qwen3.8-27B<\/td>\n27.8B<\/td>\n8GB<\/td>\napache-2.0<\/td>\nQwen3.8-27B Vision-Language Model: Our Test Answers, 8GB+ VRAM<\/a> (2026-09-26)<\/td>\n<\/tr>\n\nbartowski\/vectionlabs_Salience-27B-R6-GGUF<\/td>\n27.8B<\/td>\n12GB<\/td>\napache-2.0<\/td>\nvectionlabs_Salience-27B-R6-GGUF: Our Test Answers, 12GB+ VRAM<\/a> (2026-09-15)<\/td>\n<\/tr>\n\nagentionai\/Signal-3.8-27B-GGUF<\/td>\n27.8B<\/td>\n16GB<\/td>\napache-2.0<\/td>\nSignal-3.8-27B-GGUF: Our Test Answers, 16GB+ VRAM<\/a> (2026-09-12)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/p>\nHow to Get It<\/h2>\nThis model is distributed in GGUF format and can be run using inference engines such as llama.cpp<\/code>.<\/p>\nWhen launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
Furthermore, because this model allows dynamic specification of choices (labels) per request, task-specific pre-training or fixed labeling is unnecessary. It can process up to 52 choices in a single request during a single forward pass, and can simultaneously handle multiple questions (e.g., routing, sentiment analysis, urgency) in one request. A major strength is its support for multiple languages including Japanese (English, German, French, Hindi, Chinese, and Japanese).<\/p>\n
In a previous article introduced on our site, Decision-Making Specialized Model “OpenJev” Testing GGUF Model Released<\/a>, we covered “tinyopenjev-for-testing-gguf”, also released by ggml-org. However, that was an ultra-small test-only model (approx. 34M parameters) created solely to verify whether inference engines and loaders function properly, and its output content carried no meaning.<\/p>\nIn contrast, the newly released “OpenJev-GGUF” is a practical quantized model that retains the entire structure of the original model openjev\/openjev<\/code>. Unlike the test model, it can be used directly for actual decision-making tasks and agent inference purposes.<\/p>\n<\/p>\nHardware Requirements<\/h2>\nEstimated requirements (calculated by Local Model Watch)<\/strong> \u2014 27.4B parameters (taken from the base model openjev\/openjev)<\/p>\n\n\n\n\nYour VRAM<\/th>\nQuantization<\/th>\nFile size<\/th>\nEst. memory needed<\/th>\n<\/tr>\n<\/thead>\n\n\n24GB (RTX 4090 \/ 3090, etc.)<\/td>\nQ4_K_M<\/td>\n17.7GB<\/td>\n21.2GB<\/td>\n<\/tr>\n\n32GB (RTX 5090, etc.)<\/td>\nQ8_0<\/td>\n26.6GB<\/td>\n32.0GB<\/td>\n<\/tr>\n\n80GB class (A100 \/ H100)<\/td>\nBF16<\/td>\n50.1GB<\/td>\n60.1GB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nMemory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model’s authors. Compare with other models in our VRAM quick reference<\/a>. What the quantization names mean: glossary<\/a>.<\/em><\/p>\n<\/p>\n<\/p>\nCan You Run It Locally?<\/h2>\nRuns in Ollama, LM Studio and llama.cpp as-is.<\/strong><\/p>\nIt is distributed in GGUF, so no conversion is needed.<\/p>\nLicense \u2014 cc-by-nc-4.0<\/code> (Commercial use prohibited):<\/strong> Commercial use is prohibited<\/strong> (NC = NonCommercial). Internal business use can also count as commercial, so avoid this license for work use.<\/p>\nCompression:<\/strong> the Q4_K_M build measures 5.56 bits per weight \u2014 about 35% the size of the original 16-bit weights, calculated by this site from the actual file sizes.<\/p>\nCompiled by this site’s code from the published formats, converted builds we have found, and each engine’s own model registry. “Not found” means we have not seen such a build, not that none exists. License summaries are not legal advice \u2014 check the publisher’s original terms before relying on them.<\/em><\/p>\n<\/p>\n<\/p>\nOur Own Measurements<\/h2>\nMeasurements We Did Not Take<\/h3>\nThis model’s license (cc-by-nc-4.0<\/code>) does not permit commercial use. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<\/p>\n<\/p>\nRecent Models in the Same Size Class<\/h2>\nModels with <\/em>15\u201340B<\/em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site’s estimates; licenses are as stated on the model cards.<\/em><\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nModel<\/th>\nParameters<\/th>\nSmallest VRAM tier<\/th>\nLicense<\/th>\nOur article<\/th>\n<\/tr>\n<\/thead>\n\n\nQwen\/Qwen3.8-27B<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Vision-Language Model: 80GB+ VRAM<\/a> (2026-10-02)<\/td>\n<\/tr>\n\nCloudflare\/clef<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Structured Decision-Making Model: 80GB+ VRAM<\/a> (2026-10-01)<\/td>\n<\/tr>\n\nQwen\/Qwen3.8-27B<\/td>\n27.8B<\/td>\n8GB<\/td>\napache-2.0<\/td>\nQwen3.8-27B Vision-Language Model: Our Test Answers, 8GB+ VRAM<\/a> (2026-09-26)<\/td>\n<\/tr>\n\nbartowski\/vectionlabs_Salience-27B-R6-GGUF<\/td>\n27.8B<\/td>\n12GB<\/td>\napache-2.0<\/td>\nvectionlabs_Salience-27B-R6-GGUF: Our Test Answers, 12GB+ VRAM<\/a> (2026-09-15)<\/td>\n<\/tr>\n\nagentionai\/Signal-3.8-27B-GGUF<\/td>\n27.8B<\/td>\n16GB<\/td>\napache-2.0<\/td>\nSignal-3.8-27B-GGUF: Our Test Answers, 16GB+ VRAM<\/a> (2026-09-12)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/p>\nHow to Get It<\/h2>\nThis model is distributed in GGUF format and can be run using inference engines such as llama.cpp<\/code>.<\/p>\nWhen launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
In contrast, the newly released “OpenJev-GGUF” is a practical quantized model that retains the entire structure of the original model openjev\/openjev<\/code>. Unlike the test model, it can be used directly for actual decision-making tasks and agent inference purposes.<\/p>\n<\/p>\nHardware Requirements<\/h2>\nEstimated requirements (calculated by Local Model Watch)<\/strong> \u2014 27.4B parameters (taken from the base model openjev\/openjev)<\/p>\n\n\n\n\nYour VRAM<\/th>\nQuantization<\/th>\nFile size<\/th>\nEst. memory needed<\/th>\n<\/tr>\n<\/thead>\n\n\n24GB (RTX 4090 \/ 3090, etc.)<\/td>\nQ4_K_M<\/td>\n17.7GB<\/td>\n21.2GB<\/td>\n<\/tr>\n\n32GB (RTX 5090, etc.)<\/td>\nQ8_0<\/td>\n26.6GB<\/td>\n32.0GB<\/td>\n<\/tr>\n\n80GB class (A100 \/ H100)<\/td>\nBF16<\/td>\n50.1GB<\/td>\n60.1GB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nMemory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model’s authors. Compare with other models in our VRAM quick reference<\/a>. What the quantization names mean: glossary<\/a>.<\/em><\/p>\n<\/p>\n<\/p>\nCan You Run It Locally?<\/h2>\nRuns in Ollama, LM Studio and llama.cpp as-is.<\/strong><\/p>\nIt is distributed in GGUF, so no conversion is needed.<\/p>\nLicense \u2014 cc-by-nc-4.0<\/code> (Commercial use prohibited):<\/strong> Commercial use is prohibited<\/strong> (NC = NonCommercial). Internal business use can also count as commercial, so avoid this license for work use.<\/p>\nCompression:<\/strong> the Q4_K_M build measures 5.56 bits per weight \u2014 about 35% the size of the original 16-bit weights, calculated by this site from the actual file sizes.<\/p>\nCompiled by this site’s code from the published formats, converted builds we have found, and each engine’s own model registry. “Not found” means we have not seen such a build, not that none exists. License summaries are not legal advice \u2014 check the publisher’s original terms before relying on them.<\/em><\/p>\n<\/p>\n<\/p>\nOur Own Measurements<\/h2>\nMeasurements We Did Not Take<\/h3>\nThis model’s license (cc-by-nc-4.0<\/code>) does not permit commercial use. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<\/p>\n<\/p>\nRecent Models in the Same Size Class<\/h2>\nModels with <\/em>15\u201340B<\/em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site’s estimates; licenses are as stated on the model cards.<\/em><\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nModel<\/th>\nParameters<\/th>\nSmallest VRAM tier<\/th>\nLicense<\/th>\nOur article<\/th>\n<\/tr>\n<\/thead>\n\n\nQwen\/Qwen3.8-27B<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Vision-Language Model: 80GB+ VRAM<\/a> (2026-10-02)<\/td>\n<\/tr>\n\nCloudflare\/clef<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Structured Decision-Making Model: 80GB+ VRAM<\/a> (2026-10-01)<\/td>\n<\/tr>\n\nQwen\/Qwen3.8-27B<\/td>\n27.8B<\/td>\n8GB<\/td>\napache-2.0<\/td>\nQwen3.8-27B Vision-Language Model: Our Test Answers, 8GB+ VRAM<\/a> (2026-09-26)<\/td>\n<\/tr>\n\nbartowski\/vectionlabs_Salience-27B-R6-GGUF<\/td>\n27.8B<\/td>\n12GB<\/td>\napache-2.0<\/td>\nvectionlabs_Salience-27B-R6-GGUF: Our Test Answers, 12GB+ VRAM<\/a> (2026-09-15)<\/td>\n<\/tr>\n\nagentionai\/Signal-3.8-27B-GGUF<\/td>\n27.8B<\/td>\n16GB<\/td>\napache-2.0<\/td>\nSignal-3.8-27B-GGUF: Our Test Answers, 16GB+ VRAM<\/a> (2026-09-12)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/p>\nHow to Get It<\/h2>\nThis model is distributed in GGUF format and can be run using inference engines such as llama.cpp<\/code>.<\/p>\nWhen launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
openjev\/openjev<\/code>. Unlike the test model, it can be used directly for actual decision-making tasks and agent inference purposes.<\/p>\n<\/p>\nHardware Requirements<\/h2>\nEstimated requirements (calculated by Local Model Watch)<\/strong> \u2014 27.4B parameters (taken from the base model openjev\/openjev)<\/p>\n\n\n\n\nYour VRAM<\/th>\nQuantization<\/th>\nFile size<\/th>\nEst. memory needed<\/th>\n<\/tr>\n<\/thead>\n\n\n24GB (RTX 4090 \/ 3090, etc.)<\/td>\nQ4_K_M<\/td>\n17.7GB<\/td>\n21.2GB<\/td>\n<\/tr>\n\n32GB (RTX 5090, etc.)<\/td>\nQ8_0<\/td>\n26.6GB<\/td>\n32.0GB<\/td>\n<\/tr>\n\n80GB class (A100 \/ H100)<\/td>\nBF16<\/td>\n50.1GB<\/td>\n60.1GB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nMemory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model’s authors. Compare with other models in our VRAM quick reference<\/a>. What the quantization names mean: glossary<\/a>.<\/em><\/p>\n<\/p>\n<\/p>\nCan You Run It Locally?<\/h2>\nRuns in Ollama, LM Studio and llama.cpp as-is.<\/strong><\/p>\nIt is distributed in GGUF, so no conversion is needed.<\/p>\nLicense \u2014 cc-by-nc-4.0<\/code> (Commercial use prohibited):<\/strong> Commercial use is prohibited<\/strong> (NC = NonCommercial). Internal business use can also count as commercial, so avoid this license for work use.<\/p>\nCompression:<\/strong> the Q4_K_M build measures 5.56 bits per weight \u2014 about 35% the size of the original 16-bit weights, calculated by this site from the actual file sizes.<\/p>\nCompiled by this site’s code from the published formats, converted builds we have found, and each engine’s own model registry. “Not found” means we have not seen such a build, not that none exists. License summaries are not legal advice \u2014 check the publisher’s original terms before relying on them.<\/em><\/p>\n<\/p>\n<\/p>\nOur Own Measurements<\/h2>\nMeasurements We Did Not Take<\/h3>\nThis model’s license (cc-by-nc-4.0<\/code>) does not permit commercial use. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<\/p>\n<\/p>\nRecent Models in the Same Size Class<\/h2>\nModels with <\/em>15\u201340B<\/em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site’s estimates; licenses are as stated on the model cards.<\/em><\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nModel<\/th>\nParameters<\/th>\nSmallest VRAM tier<\/th>\nLicense<\/th>\nOur article<\/th>\n<\/tr>\n<\/thead>\n\n\nQwen\/Qwen3.8-27B<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Vision-Language Model: 80GB+ VRAM<\/a> (2026-10-02)<\/td>\n<\/tr>\n\nCloudflare\/clef<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Structured Decision-Making Model: 80GB+ VRAM<\/a> (2026-10-01)<\/td>\n<\/tr>\n\nQwen\/Qwen3.8-27B<\/td>\n27.8B<\/td>\n8GB<\/td>\napache-2.0<\/td>\nQwen3.8-27B Vision-Language Model: Our Test Answers, 8GB+ VRAM<\/a> (2026-09-26)<\/td>\n<\/tr>\n\nbartowski\/vectionlabs_Salience-27B-R6-GGUF<\/td>\n27.8B<\/td>\n12GB<\/td>\napache-2.0<\/td>\nvectionlabs_Salience-27B-R6-GGUF: Our Test Answers, 12GB+ VRAM<\/a> (2026-09-15)<\/td>\n<\/tr>\n\nagentionai\/Signal-3.8-27B-GGUF<\/td>\n27.8B<\/td>\n16GB<\/td>\napache-2.0<\/td>\nSignal-3.8-27B-GGUF: Our Test Answers, 16GB+ VRAM<\/a> (2026-09-12)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/p>\nHow to Get It<\/h2>\nThis model is distributed in GGUF format and can be run using inference engines such as llama.cpp<\/code>.<\/p>\nWhen launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
Estimated requirements (calculated by Local Model Watch)<\/strong> \u2014 27.4B parameters (taken from the base model openjev\/openjev)<\/p>\n\n\n\n\nYour VRAM<\/th>\nQuantization<\/th>\nFile size<\/th>\nEst. memory needed<\/th>\n<\/tr>\n<\/thead>\n\n\n24GB (RTX 4090 \/ 3090, etc.)<\/td>\nQ4_K_M<\/td>\n17.7GB<\/td>\n21.2GB<\/td>\n<\/tr>\n\n32GB (RTX 5090, etc.)<\/td>\nQ8_0<\/td>\n26.6GB<\/td>\n32.0GB<\/td>\n<\/tr>\n\n80GB class (A100 \/ H100)<\/td>\nBF16<\/td>\n50.1GB<\/td>\n60.1GB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\nMemory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model’s authors. Compare with other models in our VRAM quick reference<\/a>. What the quantization names mean: glossary<\/a>.<\/em><\/p>\n<\/p>\n<\/p>\nCan You Run It Locally?<\/h2>\nRuns in Ollama, LM Studio and llama.cpp as-is.<\/strong><\/p>\nIt is distributed in GGUF, so no conversion is needed.<\/p>\nLicense \u2014 cc-by-nc-4.0<\/code> (Commercial use prohibited):<\/strong> Commercial use is prohibited<\/strong> (NC = NonCommercial). Internal business use can also count as commercial, so avoid this license for work use.<\/p>\nCompression:<\/strong> the Q4_K_M build measures 5.56 bits per weight \u2014 about 35% the size of the original 16-bit weights, calculated by this site from the actual file sizes.<\/p>\nCompiled by this site’s code from the published formats, converted builds we have found, and each engine’s own model registry. “Not found” means we have not seen such a build, not that none exists. License summaries are not legal advice \u2014 check the publisher’s original terms before relying on them.<\/em><\/p>\n<\/p>\n<\/p>\nOur Own Measurements<\/h2>\nMeasurements We Did Not Take<\/h3>\nThis model’s license (cc-by-nc-4.0<\/code>) does not permit commercial use. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<\/p>\n<\/p>\nRecent Models in the Same Size Class<\/h2>\nModels with <\/em>15\u201340B<\/em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site’s estimates; licenses are as stated on the model cards.<\/em><\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nModel<\/th>\nParameters<\/th>\nSmallest VRAM tier<\/th>\nLicense<\/th>\nOur article<\/th>\n<\/tr>\n<\/thead>\n\n\nQwen\/Qwen3.8-27B<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Vision-Language Model: 80GB+ VRAM<\/a> (2026-10-02)<\/td>\n<\/tr>\n\nCloudflare\/clef<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Structured Decision-Making Model: 80GB+ VRAM<\/a> (2026-10-01)<\/td>\n<\/tr>\n\nQwen\/Qwen3.8-27B<\/td>\n27.8B<\/td>\n8GB<\/td>\napache-2.0<\/td>\nQwen3.8-27B Vision-Language Model: Our Test Answers, 8GB+ VRAM<\/a> (2026-09-26)<\/td>\n<\/tr>\n\nbartowski\/vectionlabs_Salience-27B-R6-GGUF<\/td>\n27.8B<\/td>\n12GB<\/td>\napache-2.0<\/td>\nvectionlabs_Salience-27B-R6-GGUF: Our Test Answers, 12GB+ VRAM<\/a> (2026-09-15)<\/td>\n<\/tr>\n\nagentionai\/Signal-3.8-27B-GGUF<\/td>\n27.8B<\/td>\n16GB<\/td>\napache-2.0<\/td>\nSignal-3.8-27B-GGUF: Our Test Answers, 16GB+ VRAM<\/a> (2026-09-12)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/p>\nHow to Get It<\/h2>\nThis model is distributed in GGUF format and can be run using inference engines such as llama.cpp<\/code>.<\/p>\nWhen launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
Memory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model’s authors. Compare with other models in our VRAM quick reference<\/a>. What the quantization names mean: glossary<\/a>.<\/em><\/p>\n<\/p>\n<\/p>\nCan You Run It Locally?<\/h2>\nRuns in Ollama, LM Studio and llama.cpp as-is.<\/strong><\/p>\nIt is distributed in GGUF, so no conversion is needed.<\/p>\nLicense \u2014 cc-by-nc-4.0<\/code> (Commercial use prohibited):<\/strong> Commercial use is prohibited<\/strong> (NC = NonCommercial). Internal business use can also count as commercial, so avoid this license for work use.<\/p>\nCompression:<\/strong> the Q4_K_M build measures 5.56 bits per weight \u2014 about 35% the size of the original 16-bit weights, calculated by this site from the actual file sizes.<\/p>\nCompiled by this site’s code from the published formats, converted builds we have found, and each engine’s own model registry. “Not found” means we have not seen such a build, not that none exists. License summaries are not legal advice \u2014 check the publisher’s original terms before relying on them.<\/em><\/p>\n<\/p>\n<\/p>\nOur Own Measurements<\/h2>\nMeasurements We Did Not Take<\/h3>\nThis model’s license (cc-by-nc-4.0<\/code>) does not permit commercial use. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<\/p>\n<\/p>\nRecent Models in the Same Size Class<\/h2>\nModels with <\/em>15\u201340B<\/em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site’s estimates; licenses are as stated on the model cards.<\/em><\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nModel<\/th>\nParameters<\/th>\nSmallest VRAM tier<\/th>\nLicense<\/th>\nOur article<\/th>\n<\/tr>\n<\/thead>\n\n\nQwen\/Qwen3.8-27B<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Vision-Language Model: 80GB+ VRAM<\/a> (2026-10-02)<\/td>\n<\/tr>\n\nCloudflare\/clef<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Structured Decision-Making Model: 80GB+ VRAM<\/a> (2026-10-01)<\/td>\n<\/tr>\n\nQwen\/Qwen3.8-27B<\/td>\n27.8B<\/td>\n8GB<\/td>\napache-2.0<\/td>\nQwen3.8-27B Vision-Language Model: Our Test Answers, 8GB+ VRAM<\/a> (2026-09-26)<\/td>\n<\/tr>\n\nbartowski\/vectionlabs_Salience-27B-R6-GGUF<\/td>\n27.8B<\/td>\n12GB<\/td>\napache-2.0<\/td>\nvectionlabs_Salience-27B-R6-GGUF: Our Test Answers, 12GB+ VRAM<\/a> (2026-09-15)<\/td>\n<\/tr>\n\nagentionai\/Signal-3.8-27B-GGUF<\/td>\n27.8B<\/td>\n16GB<\/td>\napache-2.0<\/td>\nSignal-3.8-27B-GGUF: Our Test Answers, 16GB+ VRAM<\/a> (2026-09-12)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/p>\nHow to Get It<\/h2>\nThis model is distributed in GGUF format and can be run using inference engines such as llama.cpp<\/code>.<\/p>\nWhen launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
Runs in Ollama, LM Studio and llama.cpp as-is.<\/strong><\/p>\nIt is distributed in GGUF, so no conversion is needed.<\/p>\nLicense \u2014 cc-by-nc-4.0<\/code> (Commercial use prohibited):<\/strong> Commercial use is prohibited<\/strong> (NC = NonCommercial). Internal business use can also count as commercial, so avoid this license for work use.<\/p>\nCompression:<\/strong> the Q4_K_M build measures 5.56 bits per weight \u2014 about 35% the size of the original 16-bit weights, calculated by this site from the actual file sizes.<\/p>\nCompiled by this site’s code from the published formats, converted builds we have found, and each engine’s own model registry. “Not found” means we have not seen such a build, not that none exists. License summaries are not legal advice \u2014 check the publisher’s original terms before relying on them.<\/em><\/p>\n<\/p>\n<\/p>\nOur Own Measurements<\/h2>\nMeasurements We Did Not Take<\/h3>\nThis model’s license (cc-by-nc-4.0<\/code>) does not permit commercial use. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<\/p>\n<\/p>\nRecent Models in the Same Size Class<\/h2>\nModels with <\/em>15\u201340B<\/em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site’s estimates; licenses are as stated on the model cards.<\/em><\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nModel<\/th>\nParameters<\/th>\nSmallest VRAM tier<\/th>\nLicense<\/th>\nOur article<\/th>\n<\/tr>\n<\/thead>\n\n\nQwen\/Qwen3.8-27B<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Vision-Language Model: 80GB+ VRAM<\/a> (2026-10-02)<\/td>\n<\/tr>\n\nCloudflare\/clef<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Structured Decision-Making Model: 80GB+ VRAM<\/a> (2026-10-01)<\/td>\n<\/tr>\n\nQwen\/Qwen3.8-27B<\/td>\n27.8B<\/td>\n8GB<\/td>\napache-2.0<\/td>\nQwen3.8-27B Vision-Language Model: Our Test Answers, 8GB+ VRAM<\/a> (2026-09-26)<\/td>\n<\/tr>\n\nbartowski\/vectionlabs_Salience-27B-R6-GGUF<\/td>\n27.8B<\/td>\n12GB<\/td>\napache-2.0<\/td>\nvectionlabs_Salience-27B-R6-GGUF: Our Test Answers, 12GB+ VRAM<\/a> (2026-09-15)<\/td>\n<\/tr>\n\nagentionai\/Signal-3.8-27B-GGUF<\/td>\n27.8B<\/td>\n16GB<\/td>\napache-2.0<\/td>\nSignal-3.8-27B-GGUF: Our Test Answers, 16GB+ VRAM<\/a> (2026-09-12)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/p>\nHow to Get It<\/h2>\nThis model is distributed in GGUF format and can be run using inference engines such as llama.cpp<\/code>.<\/p>\nWhen launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
It is distributed in GGUF, so no conversion is needed.<\/p>\n
License \u2014 cc-by-nc-4.0<\/code> (Commercial use prohibited):<\/strong> Commercial use is prohibited<\/strong> (NC = NonCommercial). Internal business use can also count as commercial, so avoid this license for work use.<\/p>\nCompression:<\/strong> the Q4_K_M build measures 5.56 bits per weight \u2014 about 35% the size of the original 16-bit weights, calculated by this site from the actual file sizes.<\/p>\nCompiled by this site’s code from the published formats, converted builds we have found, and each engine’s own model registry. “Not found” means we have not seen such a build, not that none exists. License summaries are not legal advice \u2014 check the publisher’s original terms before relying on them.<\/em><\/p>\n<\/p>\n<\/p>\nOur Own Measurements<\/h2>\nMeasurements We Did Not Take<\/h3>\nThis model’s license (cc-by-nc-4.0<\/code>) does not permit commercial use. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<\/p>\n<\/p>\nRecent Models in the Same Size Class<\/h2>\nModels with <\/em>15\u201340B<\/em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site’s estimates; licenses are as stated on the model cards.<\/em><\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nModel<\/th>\nParameters<\/th>\nSmallest VRAM tier<\/th>\nLicense<\/th>\nOur article<\/th>\n<\/tr>\n<\/thead>\n\n\nQwen\/Qwen3.8-27B<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Vision-Language Model: 80GB+ VRAM<\/a> (2026-10-02)<\/td>\n<\/tr>\n\nCloudflare\/clef<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Structured Decision-Making Model: 80GB+ VRAM<\/a> (2026-10-01)<\/td>\n<\/tr>\n\nQwen\/Qwen3.8-27B<\/td>\n27.8B<\/td>\n8GB<\/td>\napache-2.0<\/td>\nQwen3.8-27B Vision-Language Model: Our Test Answers, 8GB+ VRAM<\/a> (2026-09-26)<\/td>\n<\/tr>\n\nbartowski\/vectionlabs_Salience-27B-R6-GGUF<\/td>\n27.8B<\/td>\n12GB<\/td>\napache-2.0<\/td>\nvectionlabs_Salience-27B-R6-GGUF: Our Test Answers, 12GB+ VRAM<\/a> (2026-09-15)<\/td>\n<\/tr>\n\nagentionai\/Signal-3.8-27B-GGUF<\/td>\n27.8B<\/td>\n16GB<\/td>\napache-2.0<\/td>\nSignal-3.8-27B-GGUF: Our Test Answers, 16GB+ VRAM<\/a> (2026-09-12)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/p>\nHow to Get It<\/h2>\nThis model is distributed in GGUF format and can be run using inference engines such as llama.cpp<\/code>.<\/p>\nWhen launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
cc-by-nc-4.0<\/code> (Commercial use prohibited):<\/strong> Commercial use is prohibited<\/strong> (NC = NonCommercial). Internal business use can also count as commercial, so avoid this license for work use.<\/p>\nCompression:<\/strong> the Q4_K_M build measures 5.56 bits per weight \u2014 about 35% the size of the original 16-bit weights, calculated by this site from the actual file sizes.<\/p>\nCompiled by this site’s code from the published formats, converted builds we have found, and each engine’s own model registry. “Not found” means we have not seen such a build, not that none exists. License summaries are not legal advice \u2014 check the publisher’s original terms before relying on them.<\/em><\/p>\n<\/p>\n<\/p>\nOur Own Measurements<\/h2>\nMeasurements We Did Not Take<\/h3>\nThis model’s license (cc-by-nc-4.0<\/code>) does not permit commercial use. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<\/p>\n<\/p>\nRecent Models in the Same Size Class<\/h2>\nModels with <\/em>15\u201340B<\/em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site’s estimates; licenses are as stated on the model cards.<\/em><\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nModel<\/th>\nParameters<\/th>\nSmallest VRAM tier<\/th>\nLicense<\/th>\nOur article<\/th>\n<\/tr>\n<\/thead>\n\n\nQwen\/Qwen3.8-27B<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Vision-Language Model: 80GB+ VRAM<\/a> (2026-10-02)<\/td>\n<\/tr>\n\nCloudflare\/clef<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Structured Decision-Making Model: 80GB+ VRAM<\/a> (2026-10-01)<\/td>\n<\/tr>\n\nQwen\/Qwen3.8-27B<\/td>\n27.8B<\/td>\n8GB<\/td>\napache-2.0<\/td>\nQwen3.8-27B Vision-Language Model: Our Test Answers, 8GB+ VRAM<\/a> (2026-09-26)<\/td>\n<\/tr>\n\nbartowski\/vectionlabs_Salience-27B-R6-GGUF<\/td>\n27.8B<\/td>\n12GB<\/td>\napache-2.0<\/td>\nvectionlabs_Salience-27B-R6-GGUF: Our Test Answers, 12GB+ VRAM<\/a> (2026-09-15)<\/td>\n<\/tr>\n\nagentionai\/Signal-3.8-27B-GGUF<\/td>\n27.8B<\/td>\n16GB<\/td>\napache-2.0<\/td>\nSignal-3.8-27B-GGUF: Our Test Answers, 16GB+ VRAM<\/a> (2026-09-12)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/p>\nHow to Get It<\/h2>\nThis model is distributed in GGUF format and can be run using inference engines such as llama.cpp<\/code>.<\/p>\nWhen launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
Compression:<\/strong> the Q4_K_M build measures 5.56 bits per weight \u2014 about 35% the size of the original 16-bit weights, calculated by this site from the actual file sizes.<\/p>\nCompiled by this site’s code from the published formats, converted builds we have found, and each engine’s own model registry. “Not found” means we have not seen such a build, not that none exists. License summaries are not legal advice \u2014 check the publisher’s original terms before relying on them.<\/em><\/p>\n<\/p>\n<\/p>\nOur Own Measurements<\/h2>\nMeasurements We Did Not Take<\/h3>\nThis model’s license (cc-by-nc-4.0<\/code>) does not permit commercial use. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<\/p>\n<\/p>\nRecent Models in the Same Size Class<\/h2>\nModels with <\/em>15\u201340B<\/em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site’s estimates; licenses are as stated on the model cards.<\/em><\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nModel<\/th>\nParameters<\/th>\nSmallest VRAM tier<\/th>\nLicense<\/th>\nOur article<\/th>\n<\/tr>\n<\/thead>\n\n\nQwen\/Qwen3.8-27B<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Vision-Language Model: 80GB+ VRAM<\/a> (2026-10-02)<\/td>\n<\/tr>\n\nCloudflare\/clef<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Structured Decision-Making Model: 80GB+ VRAM<\/a> (2026-10-01)<\/td>\n<\/tr>\n\nQwen\/Qwen3.8-27B<\/td>\n27.8B<\/td>\n8GB<\/td>\napache-2.0<\/td>\nQwen3.8-27B Vision-Language Model: Our Test Answers, 8GB+ VRAM<\/a> (2026-09-26)<\/td>\n<\/tr>\n\nbartowski\/vectionlabs_Salience-27B-R6-GGUF<\/td>\n27.8B<\/td>\n12GB<\/td>\napache-2.0<\/td>\nvectionlabs_Salience-27B-R6-GGUF: Our Test Answers, 12GB+ VRAM<\/a> (2026-09-15)<\/td>\n<\/tr>\n\nagentionai\/Signal-3.8-27B-GGUF<\/td>\n27.8B<\/td>\n16GB<\/td>\napache-2.0<\/td>\nSignal-3.8-27B-GGUF: Our Test Answers, 16GB+ VRAM<\/a> (2026-09-12)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/p>\nHow to Get It<\/h2>\nThis model is distributed in GGUF format and can be run using inference engines such as llama.cpp<\/code>.<\/p>\nWhen launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
Compiled by this site’s code from the published formats, converted builds we have found, and each engine’s own model registry. “Not found” means we have not seen such a build, not that none exists. License summaries are not legal advice \u2014 check the publisher’s original terms before relying on them.<\/em><\/p>\n<\/p>\n<\/p>\nOur Own Measurements<\/h2>\nMeasurements We Did Not Take<\/h3>\nThis model’s license (cc-by-nc-4.0<\/code>) does not permit commercial use. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<\/p>\n<\/p>\nRecent Models in the Same Size Class<\/h2>\nModels with <\/em>15\u201340B<\/em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site’s estimates; licenses are as stated on the model cards.<\/em><\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nModel<\/th>\nParameters<\/th>\nSmallest VRAM tier<\/th>\nLicense<\/th>\nOur article<\/th>\n<\/tr>\n<\/thead>\n\n\nQwen\/Qwen3.8-27B<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Vision-Language Model: 80GB+ VRAM<\/a> (2026-10-02)<\/td>\n<\/tr>\n\nCloudflare\/clef<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Structured Decision-Making Model: 80GB+ VRAM<\/a> (2026-10-01)<\/td>\n<\/tr>\n\nQwen\/Qwen3.8-27B<\/td>\n27.8B<\/td>\n8GB<\/td>\napache-2.0<\/td>\nQwen3.8-27B Vision-Language Model: Our Test Answers, 8GB+ VRAM<\/a> (2026-09-26)<\/td>\n<\/tr>\n\nbartowski\/vectionlabs_Salience-27B-R6-GGUF<\/td>\n27.8B<\/td>\n12GB<\/td>\napache-2.0<\/td>\nvectionlabs_Salience-27B-R6-GGUF: Our Test Answers, 12GB+ VRAM<\/a> (2026-09-15)<\/td>\n<\/tr>\n\nagentionai\/Signal-3.8-27B-GGUF<\/td>\n27.8B<\/td>\n16GB<\/td>\napache-2.0<\/td>\nSignal-3.8-27B-GGUF: Our Test Answers, 16GB+ VRAM<\/a> (2026-09-12)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/p>\nHow to Get It<\/h2>\nThis model is distributed in GGUF format and can be run using inference engines such as llama.cpp<\/code>.<\/p>\nWhen launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
This model’s license (cc-by-nc-4.0<\/code>) does not permit commercial use. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<\/p>\n<\/p>\nRecent Models in the Same Size Class<\/h2>\nModels with <\/em>15\u201340B<\/em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site’s estimates; licenses are as stated on the model cards.<\/em><\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nModel<\/th>\nParameters<\/th>\nSmallest VRAM tier<\/th>\nLicense<\/th>\nOur article<\/th>\n<\/tr>\n<\/thead>\n\n\nQwen\/Qwen3.8-27B<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Vision-Language Model: 80GB+ VRAM<\/a> (2026-10-02)<\/td>\n<\/tr>\n\nCloudflare\/clef<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Structured Decision-Making Model: 80GB+ VRAM<\/a> (2026-10-01)<\/td>\n<\/tr>\n\nQwen\/Qwen3.8-27B<\/td>\n27.8B<\/td>\n8GB<\/td>\napache-2.0<\/td>\nQwen3.8-27B Vision-Language Model: Our Test Answers, 8GB+ VRAM<\/a> (2026-09-26)<\/td>\n<\/tr>\n\nbartowski\/vectionlabs_Salience-27B-R6-GGUF<\/td>\n27.8B<\/td>\n12GB<\/td>\napache-2.0<\/td>\nvectionlabs_Salience-27B-R6-GGUF: Our Test Answers, 12GB+ VRAM<\/a> (2026-09-15)<\/td>\n<\/tr>\n\nagentionai\/Signal-3.8-27B-GGUF<\/td>\n27.8B<\/td>\n16GB<\/td>\napache-2.0<\/td>\nSignal-3.8-27B-GGUF: Our Test Answers, 16GB+ VRAM<\/a> (2026-09-12)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/p>\nHow to Get It<\/h2>\nThis model is distributed in GGUF format and can be run using inference engines such as llama.cpp<\/code>.<\/p>\nWhen launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
cc-by-nc-4.0<\/code>) does not permit commercial use. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<\/p>\n<\/p>\nRecent Models in the Same Size Class<\/h2>\nModels with <\/em>15\u201340B<\/em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site’s estimates; licenses are as stated on the model cards.<\/em><\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nModel<\/th>\nParameters<\/th>\nSmallest VRAM tier<\/th>\nLicense<\/th>\nOur article<\/th>\n<\/tr>\n<\/thead>\n\n\nQwen\/Qwen3.8-27B<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Vision-Language Model: 80GB+ VRAM<\/a> (2026-10-02)<\/td>\n<\/tr>\n\nCloudflare\/clef<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Structured Decision-Making Model: 80GB+ VRAM<\/a> (2026-10-01)<\/td>\n<\/tr>\n\nQwen\/Qwen3.8-27B<\/td>\n27.8B<\/td>\n8GB<\/td>\napache-2.0<\/td>\nQwen3.8-27B Vision-Language Model: Our Test Answers, 8GB+ VRAM<\/a> (2026-09-26)<\/td>\n<\/tr>\n\nbartowski\/vectionlabs_Salience-27B-R6-GGUF<\/td>\n27.8B<\/td>\n12GB<\/td>\napache-2.0<\/td>\nvectionlabs_Salience-27B-R6-GGUF: Our Test Answers, 12GB+ VRAM<\/a> (2026-09-15)<\/td>\n<\/tr>\n\nagentionai\/Signal-3.8-27B-GGUF<\/td>\n27.8B<\/td>\n16GB<\/td>\napache-2.0<\/td>\nSignal-3.8-27B-GGUF: Our Test Answers, 16GB+ VRAM<\/a> (2026-09-12)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/p>\nHow to Get It<\/h2>\nThis model is distributed in GGUF format and can be run using inference engines such as llama.cpp<\/code>.<\/p>\nWhen launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
Models with <\/em>15\u201340B<\/em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site’s estimates; licenses are as stated on the model cards.<\/em><\/p>\n\n\u2192 Scroll horizontally to see all columns<\/p>\n\n\n\nModel<\/th>\nParameters<\/th>\nSmallest VRAM tier<\/th>\nLicense<\/th>\nOur article<\/th>\n<\/tr>\n<\/thead>\n\n\nQwen\/Qwen3.8-27B<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Vision-Language Model: 80GB+ VRAM<\/a> (2026-10-02)<\/td>\n<\/tr>\n\nCloudflare\/clef<\/td>\n27.4B<\/td>\n80GB<\/td>\napache-2.0<\/td>\nclef Structured Decision-Making Model: 80GB+ VRAM<\/a> (2026-10-01)<\/td>\n<\/tr>\n\nQwen\/Qwen3.8-27B<\/td>\n27.8B<\/td>\n8GB<\/td>\napache-2.0<\/td>\nQwen3.8-27B Vision-Language Model: Our Test Answers, 8GB+ VRAM<\/a> (2026-09-26)<\/td>\n<\/tr>\n\nbartowski\/vectionlabs_Salience-27B-R6-GGUF<\/td>\n27.8B<\/td>\n12GB<\/td>\napache-2.0<\/td>\nvectionlabs_Salience-27B-R6-GGUF: Our Test Answers, 12GB+ VRAM<\/a> (2026-09-15)<\/td>\n<\/tr>\n\nagentionai\/Signal-3.8-27B-GGUF<\/td>\n27.8B<\/td>\n16GB<\/td>\napache-2.0<\/td>\nSignal-3.8-27B-GGUF: Our Test Answers, 16GB+ VRAM<\/a> (2026-09-12)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<\/p>\nHow to Get It<\/h2>\nThis model is distributed in GGUF format and can be run using inference engines such as llama.cpp<\/code>.<\/p>\nWhen launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
This model is distributed in GGUF format and can be run using inference engines such as llama.cpp<\/code>.<\/p>\nWhen launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
llama.cpp<\/code>.<\/p>\nWhen launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
When launching, you can use the llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
llama serve<\/code> command included in llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
llama.cpp<\/code> to load the model directly from the Hugging Face repository and start a server. Running the following command starts the API server in your local environment:<\/p>\nllama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
llama serve -hf ggml-org\/OpenJev-GGUF\n<\/code><\/pre>\nAfter startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
After startup, you can send decision-making requests via the dedicated API endpoint \/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
\/v1\/systemone<\/code>.<\/p>\nNote that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n<\/p>\nRelated Articles<\/h2>\n\nTesting OpenJev GGUF: A Tiny Model for Runtime Loaders<\/a><\/li>\nMiMo-V2.6-Flash-RL-GGUF Multimodal MoE Model: ~141GB Memory<\/a><\/li>\nMiMo-V2.6-Distill-Qwen-9B-GGUF: Our Test Answers, 12GB+ VRAM<\/a><\/li>\nggml v0.24.0 Released with Backend Improvements and API Updates<\/a><\/li>\n<\/ul>\n<\/p>\n<\/p>\nWhat to Read Next<\/h2>\n\nFind models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 Other models that run on a 24GB GPU<\/a><\/li>\nEngines that run this model<\/strong> \u2192 llama.cpp<\/a> \/ Ollama<\/a><\/li>\nWhat Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 GGUF format guide and models<\/a> \/ Quantization and model-format glossary<\/a><\/li>\nOther models for the same task<\/strong> \u2192 Other vision-language models<\/a><\/li>\n<\/ul>\n<\/p>\nSources<\/h2>\n\nhttps:\/\/huggingface.co\/ggml-org\/OpenJev-GGUF<\/a><\/li>\nhttps:\/\/huggingface.co\/openjev\/openjev<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
Note that the weights of this model are released under the “CC BY-NC 4.0” license. While it is free for non-commercial research and use, acquiring a separate commercial license is required for commercial use.<\/p>\n
Explore OpenJev-GGUF, a GGUF quantized open-weight decision-making model by ggml-org designed for structured choices and agent tasks.<\/p>\n","protected":false},"author":1,"featured_media":8895,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[2856,161,1970,163,1073,2818,1547,1952],"class_list":["post-8896","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-decision-model-en","tag-ggml-en","tag-ggml-org-en","tag-gguf-en","tag-llama-cpp-en","tag-openjev-en","tag-verified","tag-vlm-en"],"lang":"en","translations":{"en":8896,"ja":8894},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=8896"}],"version-history":[{"count":1,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions"}],"predecessor-version":[{"id":9022,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/8896\/revisions\/9022"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/8895"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=8896"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=8896"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=8896"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}