{"id":14351,"date":"2026-11-17T11:03:04","date_gmt":"2026-11-17T02:03:04","guid":{"rendered":"https:\/\/localmodelwatch.tsuchitsuchi.com\/?p=14351"},"modified":"2026-11-17T11:12:26","modified_gmt":"2026-11-17T02:12:26","slug":"zwz2-27b-gguf-released","status":"publish","type":"post","link":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/zwz2-27b-gguf-released\/","title":{"rendered":"ZwZ2-27B-GGUF Vision-Language Model: 24GB+ VRAM"},"content":{"rendered":"<p><!-- lmw:facts --><\/p>\n<h2>At a Glance<\/h2>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>Item<\/th>\n<th>Value<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>Repository<\/td>\n<td><a href=\"https:\/\/huggingface.co\/inclusionAI\/ZwZ2-27B-GGUF\">inclusionAI\/ZwZ2-27B-GGUF<\/a><\/td>\n<\/tr>\n<tr>\n<td>Publisher guide<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/publisher-inclusionai-en\/\">inclusionAI (Ant Group): models and licenses<\/a><\/td>\n<\/tr>\n<tr>\n<td>Published<\/td>\n<td>2026-10-30<\/td>\n<\/tr>\n<tr>\n<td>Formats<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/format-gguf-en\/\">GGUF<\/a><\/td>\n<\/tr>\n<tr>\n<td>Paper<\/td>\n<td><a href=\"https:\/\/arxiv.org\/abs\/2610.37037\">arXiv:2610.37037<\/a><\/td>\n<\/tr>\n<tr>\n<td>Source type<\/td>\n<td>Primary source (the publisher itself)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><em>Values determined by this site&#8217;s code when the information was collected. Dates are JST.<\/em><\/p>\n<p><!-- \/lmw:facts --><\/p>\n<h2>Overview<\/h2>\n<p>-inclusionAI has released &#8220;inclusionAI\/ZwZ2-27B-GGUF&#8221;, a GGUF format version of the vision-language multimodal model &#8220;ZwZ2-27B&#8221;. This model is provided as part of the &#8220;Zooming without Zooming v2 (ZwZ2)&#8221; project, and is a 27B-scale open-weight model built on the Qwen3.6-27B base model. By using GGUF-compatible inference engines such as llama.cpp, users can run multimodal inference on their own PC or local environment.<\/p>\n<p>According to the paper &#8220;<a href=\"https:\/\/arxiv.org\/abs\/2610.37037\">Zooming without Zooming v2: From Seeing Details to Connecting Evidence for Generalizable Multimodal Perception<\/a>&#8220;, this model aims to improve spatial perception and long-context understanding, which serve as the foundation for Embodied AI and multimodal agents. The development team focused on the ability of &#8220;connecting evidence,&#8221; which identifies related visual facts, ties them to the correct targets, restores their relationships, and leads to answers. It is equipped with the ability to compare, collate, and aggregate visual evidence spanning multiple regions from the entire image and perform multi-step reasoning without using an external zoom tool during inference.<\/p>\n<h2>Specifications<\/h2>\n<ul>\n<li>Parameters: 27.4B<\/li>\n<li>Architecture: Qwen3_5ForConditionalGeneration (qwen3_5)<\/li>\n<li>Base Model: inclusionAI\/ZwZ2-27B (The original base model is Qwen3.6-27B)<\/li>\n<\/ul>\n<h2>Performance<\/h2>\n<p>This model is a quantized version converted into GGUF format, and the publishers have noted that the quantization process may affect inference accuracy. Benchmark measurement scores are based on the results published for the original model, &#8220;inclusionAI\/ZwZ2-27B&#8221;.<\/p>\n<p>According to the original model card and paper, it is reported that ZwZ2-27B outperformed the base model Qwen3.6-27B across all 25 benchmarks spanning 6 domains. The overall average accuracy (Overall Accuracy) improved by 4.99 points from the base model&#8217;s 55.52% to 60.51%. This performance improvement is explained to extend from fine-grained perception and spatial reasoning to long-context understanding, multimodal agents, and general visual reasoning.<\/p>\n<p>Detailed numerical tables for each benchmark are not provided as text in the main model card body, but figures summarizing the representative 8 benchmark highlights and the overall results of all 25 benchmarks have been published. Evaluation tasks include &#8220;MMLongBench*&#8221; using visual cores consisting of a total of 25 non-count tasks such as V-RAG, Image MM-NIAH, and VH-Multi, showing results that consistently outperform the base model in domain-specific averages and overall averages.<\/p>\n<p>These measurement results are figures published by the publisher inclusionAI itself and are not independent verification results by a third party. Furthermore, as for the training methodology, it is reported that the model used a multi-region question dataset of approximately 20,000 cases automatically generated from SA-1B images through the &#8220;EvidenceConnect&#8221; framework, and was optimized solely via reinforcement learning using GRPO (Group Relative Policy Optimization) without additional supervised fine-tuning (SFT). When running the GGUF format in a local environment, attention must be paid to the fact that some degradation or variation from the published accuracy of the original model may occur depending on the selected quantization precision (FP16, Q8_0, Q4_K_M, etc.).<\/p>\n<h2>Strengths and Use Cases<\/h2>\n<p>This model (and its original model) excels at comparing, aggregating, and collating visual evidence spanning multiple regions within an image to perform multi-step reasoning. Specifically, it is suited for the following tasks and uses:<\/p>\n<ul>\n<li><strong>Spatial Perception and Spatial Reasoning<\/strong>: The ability to capture positional relationships and spatial features among objects in an image.<\/li>\n<li><strong>Multimodal Long-Context Understanding<\/strong>: Understanding visual information with long contexts.<\/li>\n<li><strong>Multimodal Agents<\/strong>: Usage as an agent that executes tasks based on visual information.<\/li>\n<li><strong>General Visual Reasoning<\/strong>: From fine-grained perception to complex visual logic construction.<\/li>\n<\/ul>\n<p>The model&#8217;s tags include <code>vision-language<\/code>, <code>image-text-to-text<\/code>, <code>visual-reasoning<\/code>, <code>conversational<\/code>, and others, assuming interactive tasks combining images and text. It is also designed to read information from the entire image without requiring external zoom tools during inference.<\/p>\n<p><!-- lmw:hardware --><\/p>\n<h2>Hardware Requirements<\/h2>\n<p><strong>Estimated requirements (calculated by Local Model Watch)<\/strong> \u2014 27.4B parameters (taken from the base model inclusionAI\/ZwZ2-27B)<\/p>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>Your VRAM<\/th>\n<th>Quantization<\/th>\n<th>File size<\/th>\n<th>Est. memory needed<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>24GB (RTX 4090 \/ 3090, etc.)<\/td>\n<td>Q4_K_M<\/td>\n<td>15.4GB<\/td>\n<td>18.5GB<\/td>\n<\/tr>\n<tr>\n<td>32GB (RTX 5090, etc.)<\/td>\n<td>Q8_0<\/td>\n<td>26.6GB<\/td>\n<td>32.0GB<\/td>\n<\/tr>\n<tr>\n<td>80GB class (A100 \/ H100)<\/td>\n<td>F16<\/td>\n<td>50.1GB<\/td>\n<td>60.1GB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><em>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&#8217;s authors. Compare with other models in our <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/vram-guide-en\/\">VRAM quick reference<\/a>. What the quantization names mean: <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/glossary-quantization-en\/\">glossary<\/a>.<\/em><\/p>\n<p><!-- \/lmw:hardware --><\/p>\n<p><!-- lmw:runnability --><\/p>\n<h2>Can You Run It Locally?<\/h2>\n<p><strong>Runs in Ollama, LM Studio and llama.cpp as-is.<\/strong><\/p>\n<p>It is distributed in GGUF, so no conversion is needed.<\/p>\n<p><strong>Compression:<\/strong> the Q4_K_M build measures 4.84 bits per weight \u2014 about 30% the size of the original 16-bit weights, calculated by this site from the actual file sizes.<\/p>\n<p><em>Compiled by this site&#8217;s code from the published formats, converted builds we have found, and each engine&#8217;s own model registry. &#8220;Not found&#8221; means we have not seen such a build, not that none exists.<\/em><\/p>\n<p><!-- \/lmw:runnability --><\/p>\n<p><!-- lmw:lab --><\/p>\n<h2>Our Own Measurements<\/h2>\n<h3>Measurements We Did Not Take<\/h3>\n<p>This model does not state a license, so we cannot confirm permission to use it. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<p><!-- \/lmw:lab --><\/p>\n<p><!-- lmw:peers --><\/p>\n<h2>Recent Models in the Same Size Class<\/h2>\n<p><em>Models with <\/em><em>15\u201340B<\/em><em> parameters that Local Model Watch covered recently, listed by code from our article log for comparison. VRAM tiers are this site&#8217;s estimates; licenses are as stated on the model cards.<\/em><\/p>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<p class=\"lmw-table-hint\" style=\"margin:0 0 4px;font-size:0.85em;opacity:0.7;\">\u2192 Scroll horizontally to see all columns<\/p>\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>Model<\/th>\n<th>Parameters<\/th>\n<th>Smallest VRAM tier<\/th>\n<th>License<\/th>\n<th>Our article<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>ggml-org\/Muse-Glimmer-30B-GGUF<\/td>\n<td>29.8B<\/td>\n<td>24GB<\/td>\n<td>apache-2.0<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/muse-glimmer-30b-gguf-2\/\">Muse-Glimmer-30B-GGUF 30B Multimodal Local Agent Model: 24GB+ VRAM<\/a> (2026-11-17)<\/td>\n<\/tr>\n<tr>\n<td>Cloudflare\/clef-omni<\/td>\n<td>35.3B<\/td>\n<td>80GB<\/td>\n<td>apache-2.0<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/cloudflare-releases-clef-omni\/\">clef-omni 30B Multimodal Decision Model: 80GB+ VRAM<\/a> (2026-11-17)<\/td>\n<\/tr>\n<tr>\n<td>ggml-org\/vllm-d3-GGUF<\/td>\n<td>26.1B<\/td>\n<td>24GB<\/td>\n<td>apache-2.0<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/vllm-d3-gguf-models-released\/\">vllm-d3 GGUF Models Released: Decision Models for Local Run<\/a> (2026-11-17)<\/td>\n<\/tr>\n<tr>\n<td>Cloudflare\/clef<\/td>\n<td>27.4B<\/td>\n<td>12GB<\/td>\n<td>apache-2.0<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/10\/02\/cloudflare-clef-27b-multimodal-decision-model\/\">clef Structured Decision-Making Model: 12GB+ VRAM, GGUF Builds<\/a> (2026-10-01)<\/td>\n<\/tr>\n<tr>\n<td>ggml-org\/OpenJev-GGUF<\/td>\n<td>27.4B<\/td>\n<td>24GB<\/td>\n<td>cc-by-nc-4.0<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/10\/02\/ggml-org-openjev-gguf-released-2\/\">OpenJev-GGUF 27.4B Decision-Making Specialized Model: 24GB+ VRAM<\/a> (2026-10-01)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><!-- \/lmw:peers --><\/p>\n<h2>How to Get It<\/h2>\n<p>This model is distributed in GGUF format and is available in the following formats:<\/p>\n<ul>\n<li><strong>Distribution Format<\/strong>: GGUF<\/li>\n<li><strong>Supported Components<\/strong>: The language model body (Language model) and the image processing projector (Vision component \/ <code>mmproj<\/code>) are provided as separate files. Depending on the use case, different precisions such as FP16, Q8_0, and Q4_K_M can be combined and used.<\/li>\n<li><strong>Supported Engines<\/strong>: Operation is recommended on relatively recent versions of <code>llama.cpp<\/code> that feature multimodal support for Qwen3.5.<\/li>\n<\/ul>\n<h3>CLI Inference Example<\/h3>\n<pre><code class=\"language-bash\">llama-mtmd-cli \\\n  -m ZwZ2-27B-Q8_0.gguf \\\n  --mmproj mmproj-ZwZ2-27B-Q8_0.gguf \\\n  --image \/path\/to\/image.jpg \\\n  -p &quot;Describe this image.&quot; \\\n  --temp 0 -n 2048\n<\/code><\/pre>\n<h3>Local API Server Example<\/h3>\n<pre><code class=\"language-bash\">llama-server \\\n  -m ZwZ2-27B-Q8_0.gguf \\\n  --mmproj mmproj-ZwZ2-27B-Q8_0.gguf \\\n  --host 127.0.0.1 --port 8080\n<\/code><\/pre>\n<p>*Please read file names accordingly based on the precision you use (FP16, Q4_K_M, etc.). When performing image input, you must specify the corresponding <code>mmproj<\/code> component.<\/p>\n<p><!-- lmw:related --><\/p>\n<h2>Related Articles<\/h2>\n<ul>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/zoomgain-9b-gguf-released\/\">ZoomGain-9B-GGUF Vision-Language Model: 8GB+ VRAM<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/zwz2-4b-gguf-released\/\">ZwZ2-4B-GGUF Vision-Language Model: 4GB+ VRAM<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/zwz2-9b-gguf-released\/\">ZwZ2-9B-GGUF Vision-Language Model: 8GB+ VRAM<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/09\/19\/realtime-venus-multimodal-conversational-ai\/\">Realtime-Venus Full-Duplex Audio-Video Dialog Model: 24GB+ VRAM<\/a><\/li>\n<\/ul>\n<p><!-- \/lmw:related --><\/p>\n<p><!-- lmw:next-steps --><\/p>\n<h2>What to Read Next<\/h2>\n<ul>\n<li><strong>Find models by VRAM<\/strong> (This model runs from the 24GB tier) \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/vram-guide-24gb-en\/\">Other models that run on a 24GB GPU<\/a><\/li>\n<li><strong>Engines that run this model<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/engine-llama-cpp-en\/\">llama.cpp<\/a> \/ <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/engine-ollama-en\/\">Ollama<\/a><\/li>\n<li><strong>What Q4_K_M, Q8_0 mean and where to get this model<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/format-gguf-en\/\">GGUF format guide and models<\/a> \/ <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/glossary-quantization-en\/\">Quantization and model-format glossary<\/a><\/li>\n<li><strong>Learn about the publisher<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/publisher-inclusionai-en\/\">inclusionAI (Ant Group): models, licenses and articles<\/a><\/li>\n<li><strong>Other models for the same task<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/models-by-task-en\/#task-vision\">Other vision-language models<\/a><\/li>\n<\/ul>\n<p><!-- \/lmw:next-steps --><\/p>\n<h2>Sources<\/h2>\n<ul>\n<li><a href=\"https:\/\/huggingface.co\/inclusionAI\/ZwZ2-27B-GGUF\"><a href=\"https:\/\/huggingface.co\/inclusionAI\/ZwZ2-27B-GGUF\">https:\/\/huggingface.co\/inclusionAI\/ZwZ2-27B-GGUF<\/a><\/a><\/li>\n<li><a href=\"https:\/\/arxiv.org\/abs\/2610.37037\"><a href=\"https:\/\/arxiv.org\/abs\/2610.37037\">https:\/\/arxiv.org\/abs\/2610.37037<\/a><\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>inclusionAI has released ZwZ2-27B-GGUF, a GGUF version of the vision-language multimodal model ZwZ2-27B, runnable locally via llama.cpp.<\/p>\n","protected":false},"author":1,"featured_media":14350,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[163,1691,1073,4427,1547,4145,1952,4429],"class_list":["post-14351","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-gguf-en","tag-inclusionai-en","tag-llama-cpp-en","tag-qwen3-6-en","tag-verified","tag-vision-language-en","tag-vlm-en","tag-zwz2-27b-en"],"lang":"en","translations":{"en":14351,"ja":14349},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/14351","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=14351"}],"version-history":[{"count":2,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/14351\/revisions"}],"predecessor-version":[{"id":18006,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/14351\/revisions\/18006"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/14350"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=14351"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=14351"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=14351"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}