{"id":410,"date":"2026-09-10T15:18:57","date_gmt":"2026-09-10T06:18:57","guid":{"rendered":"https:\/\/localmodelwatch.tsuchitsuchi.com\/2026\/09\/10\/deepseek-v4-1-flash-released\/"},"modified":"2026-11-03T16:08:22","modified_gmt":"2026-11-03T07:08:22","slug":"deepseek-v4-1-flash-released","status":"publish","type":"post","link":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/09\/10\/deepseek-v4-1-flash-released\/","title":{"rendered":"DeepSeek-V4.1-Flash 552B Multimodal MoE Model: ~570GB Memory"},"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\/deepseek-ai\/DeepSeek-V4.1-Flash\">deepseek-ai\/DeepSeek-V4.1-Flash<\/a><\/td>\n<\/tr>\n<tr>\n<td>Family guide<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/model-deepseek-ai-deepseek-v4-1-en\/\">DeepSeek-V4.1 guide (1 articles)<\/a><\/td>\n<\/tr>\n<tr>\n<td>Publisher guide<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/publisher-deepseek-en\/\">DeepSeek: models and licenses<\/a><\/td>\n<\/tr>\n<tr>\n<td>Published<\/td>\n<td>2026-09-10<\/td>\n<\/tr>\n<tr>\n<td>License<\/td>\n<td>mit<\/td>\n<\/tr>\n<tr>\n<td>Formats<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/format-safetensors-en\/\">safetensors<\/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<p><!-- lmw:lab-summary --><\/p>\n<p><strong>What we checked ourselves<\/strong><\/p>\n<ul>\n<li>The same Japanese text takes about as many tokens as with the Qwen3 tokenizer.<\/li>\n<\/ul>\n<p>Details and conditions are in \u201cOur Own Measurements\u201d below.<\/p>\n<p><!-- \/lmw:lab-summary --><\/p>\n<h2>Overview<\/h2>\n<p>deepseek-ai has released &#8220;DeepSeek-V4.1-Flash&#8221;, a multimodal Mixture-of-Experts (MoE) model with a 552B backbone parameter count, on Hugging Face. It supports a context length of up to 1 million tokens, natively processes images and text, and generates text autoregressively.<\/p>\n<h2>Specifications<\/h2>\n<ul>\n<li>Parameter Count: 552B backbone (active parameters: 8B during prefill \/ 16B during decode)<\/li>\n<li>Architecture: MoE (Causal Encoder-Decoder, 1 shared expert, 384 routed experts, 6 routed experts activated per layer)<\/li>\n<li>Context Length: Up to 1 million tokens<\/li>\n<li>License: mit<\/li>\n<\/ul>\n<h2>Performance<\/h2>\n<p>Comparison tables with base models and frontier models as presented in the model card and technical report are shown below.<\/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 style=\"text-align: left;\">Benchmark (Metric)<\/th>\n<th style=\"text-align: center;\"># Shots<\/th>\n<th style=\"text-align: center;\">DeepSeek-V4-Flash-Base<\/th>\n<th style=\"text-align: center;\">DeepSeek-V4-Pro-Base<\/th>\n<th style=\"text-align: center;\">DeepSeek-V4.1-Flash-Base<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td style=\"text-align: left;\">Architecture<\/td>\n<td style=\"text-align: center;\">\u2014<\/td>\n<td style=\"text-align: center;\">MoE<\/td>\n<td style=\"text-align: center;\">MoE<\/td>\n<td style=\"text-align: center;\">MoE<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\"># Backbone Params<\/td>\n<td style=\"text-align: center;\">\u2014<\/td>\n<td style=\"text-align: center;\">284B<\/td>\n<td style=\"text-align: center;\">1.6T<\/td>\n<td style=\"text-align: center;\">552B<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\"># Activated Params<\/td>\n<td style=\"text-align: center;\">\u2014<\/td>\n<td style=\"text-align: center;\">13B<\/td>\n<td style=\"text-align: center;\">49B<\/td>\n<td style=\"text-align: center;\">8B \/ 16B<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\"><strong>World Knowledge<\/strong><\/td>\n<td style=\"text-align: center;\"><\/td>\n<td style=\"text-align: center;\"><\/td>\n<td style=\"text-align: center;\"><\/td>\n<td style=\"text-align: center;\"><\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">AGIEval (EM)<\/td>\n<td style=\"text-align: center;\">3\u20135-shot<\/td>\n<td style=\"text-align: center;\">83.9<\/td>\n<td style=\"text-align: center;\"><strong>84.4<\/strong><\/td>\n<td style=\"text-align: center;\">83.4<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">MMLU-Pro (EM)<\/td>\n<td style=\"text-align: center;\">5-shot<\/td>\n<td style=\"text-align: center;\">68.3<\/td>\n<td style=\"text-align: center;\">73.5<\/td>\n<td style=\"text-align: center;\"><strong>74.1<\/strong><\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">C-Eval (EM)<\/td>\n<td style=\"text-align: center;\">5-shot<\/td>\n<td style=\"text-align: center;\">92.1<\/td>\n<td style=\"text-align: center;\"><strong>93.1<\/strong><\/td>\n<td style=\"text-align: center;\">92.1<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">MultiLoKo (LLM-Judge)<\/td>\n<td style=\"text-align: center;\">5-shot<\/td>\n<td style=\"text-align: center;\">42.6<\/td>\n<td style=\"text-align: center;\"><strong>50.9<\/strong><\/td>\n<td style=\"text-align: center;\">45.5<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">SimpleQA-Verified (EM)<\/td>\n<td style=\"text-align: center;\">25-shot<\/td>\n<td style=\"text-align: center;\">30.1<\/td>\n<td style=\"text-align: center;\"><strong>55.2<\/strong><\/td>\n<td style=\"text-align: center;\">42.3<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">SuperGPQA (EM)<\/td>\n<td style=\"text-align: center;\">5-shot<\/td>\n<td style=\"text-align: center;\">46.5<\/td>\n<td style=\"text-align: center;\"><strong>53.9<\/strong><\/td>\n<td style=\"text-align: center;\">53.1<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\"><strong>Language &amp; Reasoning<\/strong><\/td>\n<td style=\"text-align: center;\"><\/td>\n<td style=\"text-align: center;\"><\/td>\n<td style=\"text-align: center;\"><\/td>\n<td style=\"text-align: center;\"><\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">BBH (EM)<\/td>\n<td style=\"text-align: center;\">3-shot<\/td>\n<td style=\"text-align: center;\">86.9<\/td>\n<td style=\"text-align: center;\"><strong>87.5<\/strong><\/td>\n<td style=\"text-align: center;\">86.1<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">BBEH (EM)<\/td>\n<td style=\"text-align: center;\">1-shot<\/td>\n<td style=\"text-align: center;\">25.4<\/td>\n<td style=\"text-align: center;\"><strong>29.8<\/strong><\/td>\n<td style=\"text-align: center;\">27.2<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">DROP (F1)<\/td>\n<td style=\"text-align: center;\">1-shot<\/td>\n<td style=\"text-align: center;\"><strong>88.6<\/strong><\/td>\n<td style=\"text-align: center;\"><strong>88.7<\/strong><\/td>\n<td style=\"text-align: center;\">87.9<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">HellaSwag (EM)<\/td>\n<td style=\"text-align: center;\">0-shot<\/td>\n<td style=\"text-align: center;\">85.7<\/td>\n<td style=\"text-align: center;\"><strong>88.0<\/strong><\/td>\n<td style=\"text-align: center;\">87.2<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\"><strong>Code &amp; Math<\/strong><\/td>\n<td style=\"text-align: center;\"><\/td>\n<td style=\"text-align: center;\"><\/td>\n<td style=\"text-align: center;\"><\/td>\n<td style=\"text-align: center;\"><\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">BigCodeBench (Pass@1)<\/td>\n<td style=\"text-align: center;\">3-shot<\/td>\n<td style=\"text-align: center;\">56.8<\/td>\n<td style=\"text-align: center;\">59.2<\/td>\n<td style=\"text-align: center;\"><strong>60.6<\/strong><\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">HumanEval (Pass@1)<\/td>\n<td style=\"text-align: center;\">0-shot<\/td>\n<td style=\"text-align: center;\">69.5<\/td>\n<td style=\"text-align: center;\">76.8<\/td>\n<td style=\"text-align: center;\"><strong>79.4<\/strong><\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">GSM8K (EM)<\/td>\n<td style=\"text-align: center;\">8-shot<\/td>\n<td style=\"text-align: center;\">90.8<\/td>\n<td style=\"text-align: center;\">92.6<\/td>\n<td style=\"text-align: center;\"><strong>93.0<\/strong><\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">MATH (EM)<\/td>\n<td style=\"text-align: center;\">4-shot<\/td>\n<td style=\"text-align: center;\">57.4<\/td>\n<td style=\"text-align: center;\"><strong>64.5<\/strong><\/td>\n<td style=\"text-align: center;\">61.1<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">MGSM (EM)<\/td>\n<td style=\"text-align: center;\">8-shot<\/td>\n<td style=\"text-align: center;\"><strong>85.7<\/strong><\/td>\n<td style=\"text-align: center;\">84.4<\/td>\n<td style=\"text-align: center;\">80.2<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\"><strong>Long Context<\/strong><\/td>\n<td style=\"text-align: center;\"><\/td>\n<td style=\"text-align: center;\"><\/td>\n<td style=\"text-align: center;\"><\/td>\n<td style=\"text-align: center;\"><\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">LongBench-V2 (EM)<\/td>\n<td style=\"text-align: center;\">1-shot<\/td>\n<td style=\"text-align: center;\">44.7<\/td>\n<td style=\"text-align: center;\"><strong>51.5<\/strong><\/td>\n<td style=\"text-align: center;\">45.2<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\"><strong>Multimodal<\/strong><\/td>\n<td style=\"text-align: center;\"><\/td>\n<td style=\"text-align: center;\"><\/td>\n<td style=\"text-align: center;\"><\/td>\n<td style=\"text-align: center;\"><\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">MMMU-Pro (EM)<\/td>\n<td style=\"text-align: center;\">4-shot<\/td>\n<td style=\"text-align: center;\">\u2014<\/td>\n<td style=\"text-align: center;\">\u2014<\/td>\n<td style=\"text-align: center;\">56.5<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">CVBench (EM)<\/td>\n<td style=\"text-align: center;\">4-shot<\/td>\n<td style=\"text-align: center;\">\u2014<\/td>\n<td style=\"text-align: center;\">\u2014<\/td>\n<td style=\"text-align: center;\">77.9<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">DocVQA (LLM-Judge)<\/td>\n<td style=\"text-align: center;\">4-shot<\/td>\n<td style=\"text-align: center;\">\u2014<\/td>\n<td style=\"text-align: center;\">\u2014<\/td>\n<td style=\"text-align: center;\">95.6<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">RefCOCO-avg (Acc@0.5)<\/td>\n<td style=\"text-align: center;\">0-shot<\/td>\n<td style=\"text-align: center;\">\u2014<\/td>\n<td style=\"text-align: center;\">\u2014<\/td>\n<td style=\"text-align: center;\">86.0<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p>Next, a restricted comparison table with frontier models (at maximum inference effort settings) is shown below.<\/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 style=\"text-align: left;\">Benchmark (Metric)<\/th>\n<th style=\"text-align: center;\">GPT-5.6 Sol<\/th>\n<th style=\"text-align: center;\">K3<\/th>\n<th style=\"text-align: center;\">DS-V4-Pro<\/th>\n<th style=\"text-align: center;\">DS-V4-Flash<\/th>\n<th style=\"text-align: center;\">DeepSeek-V4.1-Flash<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td style=\"text-align: left;\"><strong>Reasoning<\/strong><\/td>\n<td style=\"text-align: center;\"><\/td>\n<td style=\"text-align: center;\"><\/td>\n<td style=\"text-align: center;\"><\/td>\n<td style=\"text-align: center;\"><\/td>\n<td style=\"text-align: center;\"><\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">GPQA Diamond (Pass@1)<\/td>\n<td style=\"text-align: center;\"><strong>94.1<\/strong><\/td>\n<td style=\"text-align: center;\">92.9<\/td>\n<td style=\"text-align: center;\">92.4<\/td>\n<td style=\"text-align: center;\">89.9<\/td>\n<td style=\"text-align: center;\">90.9<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">Codeforces (Rating)<\/td>\n<td style=\"text-align: center;\">\u2014<\/td>\n<td style=\"text-align: center;\">\u2014<\/td>\n<td style=\"text-align: center;\">3348<\/td>\n<td style=\"text-align: center;\">3289<\/td>\n<td style=\"text-align: center;\"><strong>3471<\/strong><\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">MathArena Apex (Pass@1)<\/td>\n<td style=\"text-align: center;\">\u2014<\/td>\n<td style=\"text-align: center;\"><strong>65.6<\/strong><\/td>\n<td style=\"text-align: center;\">65.3<\/td>\n<td style=\"text-align: center;\">58.6<\/td>\n<td style=\"text-align: center;\"><strong>65.6<\/strong><\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\"><strong>Agentic<\/strong><\/td>\n<td style=\"text-align: center;\"><\/td>\n<td style=\"text-align: center;\"><\/td>\n<td style=\"text-align: center;\"><\/td>\n<td style=\"text-align: center;\"><\/td>\n<td style=\"text-align: center;\"><\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">Terminal-Bench 2.1 (Pass@1)<\/td>\n<td style=\"text-align: center;\">88.8<\/td>\n<td style=\"text-align: center;\">88.3<\/td>\n<td style=\"text-align: center;\">87.9<\/td>\n<td style=\"text-align: center;\">82.7<\/td>\n<td style=\"text-align: center;\"><strong>90.6<\/strong><\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">Terminal-Bench 3.0 (Pass@1)<\/td>\n<td style=\"text-align: center;\">34.4<\/td>\n<td style=\"text-align: center;\">17.7<\/td>\n<td style=\"text-align: center;\">11.8<\/td>\n<td style=\"text-align: center;\">7.6<\/td>\n<td style=\"text-align: center;\">30.0<\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">DeepSWE v1.1 (Resolved)<\/td>\n<td style=\"text-align: center;\">73.0<\/td>\n<td style=\"text-align: center;\">67.5<\/td>\n<td style=\"text-align: center;\">62.7<\/td>\n<td style=\"text-align: center;\">54.4<\/td>\n<td style=\"text-align: center;\"><strong>74.2<\/strong><\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">CyberGym (Pass@1)<\/td>\n<td style=\"text-align: center;\">84.5<\/td>\n<td style=\"text-align: center;\">80.0<\/td>\n<td style=\"text-align: center;\">83.3<\/td>\n<td style=\"text-align: center;\">76.7<\/td>\n<td style=\"text-align: center;\"><strong>88.1<\/strong><\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">AutomationBench (Pass@1)<\/td>\n<td style=\"text-align: center;\">45.8<\/td>\n<td style=\"text-align: center;\">46.7<\/td>\n<td style=\"text-align: center;\">43.2<\/td>\n<td style=\"text-align: center;\">37.7<\/td>\n<td style=\"text-align: center;\"><strong>54.8<\/strong><\/td>\n<\/tr>\n<tr>\n<td style=\"text-align: left;\">Agent&#8217;s Last Exam (Pass@1)<\/td>\n<td style=\"text-align: center;\">26.7<\/td>\n<td style=\"text-align: center;\">27.6<\/td>\n<td style=\"text-align: center;\">25.7<\/td>\n<td style=\"text-align: center;\">25.2<\/td>\n<td style=\"text-align: center;\"><strong>31.8<\/strong><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p>According to the evaluation results published by the authors, DeepSeek-V4.1-Flash achieves the highest scores in its scale category and among comparisons in coding fields (HumanEval, BigCodeBench), competitive programming (Codeforces), and agentic tasks (such as Terminal-Bench and DeepSWE v1.1). While it also demonstrates high scores in general knowledge and reasoning like MMLU-Pro, it falls behind DeepSeek-V4-Flash and DeepSeek-V4-Pro in some items such as long-context tasks (LongBench-V2) and multilingual math (MGSM).<\/p>\n<h2>Strengths and Use Cases<\/h2>\n<ul>\n<li>Multimodal Support: Multimodal tasks that process images and text simultaneously (DocVQA, RefCOCO, etc.)<\/li>\n<li>Coding and Agent Tasks: Extremely high performance in Python tasks involving multiple library calls, as well as terminal operation and software engineering agents<\/li>\n<li>Long-Context Processing: Supports context lengths up to 1 million tokens, improving cost efficiency through KV cache compression technologies (such as CSA2 and FP4 main KV cache)<\/li>\n<li>Inference Effort Control: Supports a &#8220;continuously controllable reasoning effort&#8221; setting that allows adjusting inference cost and accuracy using an integer from 1 to 100<\/li>\n<\/ul>\n<p><!-- lmw:hardware --><\/p>\n<h2>Hardware Requirements<\/h2>\n<p><strong>Estimated requirements (calculated by Local Model Watch)<\/strong> \u2014 763.2B parameters<\/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>More than 570GB of VRAM (multi-GPU or CPU offload required)<\/td>\n<td>FP8<\/td>\n<td>475.3GB<\/td>\n<td>570.3GB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><strong>Inference engine support<\/strong> (architecture name matched against each project&#8217;s own model registry in its source code, checked 2026-11-03): llama.cpp: registered, vLLM: registered, MLX (mlx-lm): registered.<\/p>\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>Not usable in Ollama, LM Studio and llama.cpp yet \u2014 we have found no GGUF build.<\/strong><\/p>\n<p>The publisher ships safetensors only. However, llama.cpp&#8217;s registry does list this architecture, so <strong>conversion to GGUF is possible<\/strong> and the model will run once someone publishes a converted build. 46 converted build(s) from other uploaders exist. Today it can be run with transformers or vLLM, using the memory figures in the table above.<\/p>\n<p><strong>License \u2014 <code>mit<\/code> (Commercial use allowed):<\/strong> Permits commercial use, modification and redistribution, provided the copyright notice and license text are retained.<\/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. License summaries are not legal advice \u2014 check the publisher&#8217;s original terms before relying on them.<\/em><\/p>\n<p><!-- \/lmw:runnability --><\/p>\n<p><!-- lmw:lab --><\/p>\n<h2>Our Own Measurements<\/h2>\n<p>Values we measured ourselves on our server (no GPU) by actually reading and running this model&#8217;s files \u2014 not figures copied from the model card. How we measure, and the results for every model: <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/observations-en\/\">Our Measurements<\/a>.<\/p>\n<h3>Japanese Token Efficiency<\/h3>\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>Tokenizer<\/th>\n<th>Tokens per 1,000 Japanese characters<\/th>\n<th>Ratio to the same text in English<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td><strong>This model<\/strong><\/td>\n<td><strong>705<\/strong><\/td>\n<td><strong>1.29\u00d7<\/strong><\/td>\n<\/tr>\n<tr>\n<td>Qwen3<\/td>\n<td>688<\/td>\n<td>1.26\u00d7<\/td>\n<\/tr>\n<tr>\n<td>Llama 3.2<\/td>\n<td>744<\/td>\n<td>1.36\u00d7<\/td>\n<\/tr>\n<tr>\n<td>Gemma 3<\/td>\n<td>564<\/td>\n<td>1.03\u00d7<\/td>\n<\/tr>\n<tr>\n<td>gpt-oss<\/td>\n<td>795<\/td>\n<td>1.45\u00d7<\/td>\n<\/tr>\n<tr>\n<td>LLM-jp-3<\/td>\n<td>497<\/td>\n<td>0.85\u00d7<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p>The same Japanese text takes about as many tokens as with the Qwen3 tokenizer.<\/p>\n<p>Counted with the <code>tokenizer.json<\/code> of <a href=\"https:\/\/huggingface.co\/deepseek-ai\/DeepSeek-V4.1-Flash\">deepseek-ai\/DeepSeek-V4.1-Flash<\/a> on a fixed text we wrote ourselves (876 Japanese characters across news, conversation, technical docs, a formal email, travel writing and a recipe) and its English translation. Fewer tokens mean more Japanese fits in the context window.<\/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>over 40B<\/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>Qwen\/Qwen3.8-Flash-Next<\/td>\n<td>180.0B<\/td>\n<td>\u2014<\/td>\n<td>qwen-community-1.0<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/10\/05\/qwen-3-8-flash-next-released\/\">Qwen3.8-Flash-Next Multimodal MoE Model: ~402GB Memory<\/a> (2026-10-04)<\/td>\n<\/tr>\n<tr>\n<td>ggml-org\/GLM-5.3-Flash-GGUF<\/td>\n<td>321.3B<\/td>\n<td>\u2014<\/td>\n<td>other<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/10\/04\/glm-5-3-flash-gguf-released\/\">GLM-5.3-Flash-GGUF Vision-Language Model: ~150GB Memory<\/a> (2026-10-03)<\/td>\n<\/tr>\n<tr>\n<td>bartowski\/Intern-S2-397B-GGUF<\/td>\n<td>403.4B<\/td>\n<td>\u2014<\/td>\n<td>apache-2.0<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/09\/15\/intern-s2-397b-gguf-released\/\">Intern-S2-397B-GGUF Vision-Language Model: ~102GB Memory<\/a> (2026-09-15)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><!-- \/lmw:peers --><\/p>\n<h2>How to Get It<\/h2>\n<ul>\n<li>Distribution Format: safetensors \/ transformers<\/li>\n<li>How to Get: Available for download from the Hugging Face repository (<code>deepseek-ai\/DeepSeek-V4.1-Flash<\/code>). No license agreement is required for access (gated: false).<\/li>\n<\/ul>\n<p><!-- lmw:variants --><\/p>\n<h2>Quantized and Converted Variants<\/h2>\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>Added<\/th>\n<th>Publisher<\/th>\n<th>Format<\/th>\n<th>Repository<\/th>\n<th>Smallest VRAM tier (build, est. memory)<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>2026-09-17 (7 days after publication)<\/td>\n<td>nvidia<\/td>\n<td>NVFP4<\/td>\n<td><a href=\"https:\/\/huggingface.co\/nvidia\/DeepSeek-V4.1-Flash-NVFP4\">nvidia\/DeepSeek-V4.1-Flash-NVFP4<\/a><\/td>\n<td>BF16 1690.8GB (does not fit a single consumer GPU)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p>In addition, 46 converted build(s) from other uploaders exist on Hugging Face; this site lists only builds from the model&#8217;s publisher or established quantization maintainers.<\/p>\n<p><em>This section is appended automatically by Local Model Watch when a converted build of this model appears after publication. Memory figures are estimated from the size of the distributed files. What the quantization names mean: <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/glossary-quantization-en\/\">glossary<\/a>.<\/em><\/p>\n<p><!-- \/lmw:variants --><\/p>\n<p><!-- lmw:related --><\/p>\n<h2>Related Articles<\/h2>\n<ul>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/09\/26\/deepseek-v4-pro-0813-released\/\">DeepSeek-V4-Pro-0813 Text Generation Model: ~998GB Memory, GGUF Builds<\/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 does not fit even in 80GB; its smallest build needs about 570GB) \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/vram-guide-en\/\">VRAM quick reference, including models over 80GB<\/a><\/li>\n<li><strong>Explore the same model family<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/model-deepseek-ai-deepseek-v4-1-en\/\">DeepSeek-V4.1 family overview (1 articles, 1 converted builds)<\/a><\/li>\n<li><strong>Engines that run this model<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/engine-vllm-en\/\">vLLM<\/a><\/li>\n<li><strong>What FP8 mean and where to get this model<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/format-fp4-en\/\">NVFP4 \/ MXFP4 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-deepseek-en\/\">DeepSeek: models, licenses and articles<\/a><\/li>\n<\/ul>\n<p><!-- \/lmw:next-steps --><\/p>\n<h2>Sources<\/h2>\n<ul>\n<li><a href=\"https:\/\/huggingface.co\/deepseek-ai\/DeepSeek-V4.1-Flash\">https:\/\/huggingface.co\/deepseek-ai\/DeepSeek-V4.1-Flash<\/a><\/li>\n<\/ul>\n<p><!-- lmw:updates --><\/p>\n<h2>Update History<\/h2>\n<ul>\n<li>2026-09-17: Added converted builds to \u201cQuantized and Converted Variants\u201d: nvidia\/DeepSeek-V4.1-Flash-NVFP4<\/li>\n<li>2026-09-24: Rebuilt the article: Added a &#8220;can you run it locally?&#8221; verdict from the distribution formats, converted builds, engine registries, license terms and measured compression<\/li>\n<li>2026-09-25: Changed the title to show what the article covers (VRAM requirements, file list, etc.).<\/li>\n<li>2026-09-28: Added our own measurements: Japanese token efficiency.<\/li>\n<\/ul>\n<p><!-- \/lmw:updates --><\/p>\n","protected":false},"excerpt":{"rendered":"<p>deepseek-ai has released DeepSeek-V4.1-Flash on Hugging Face, a multimodal MoE model supporting up to 1M tokens. Learn its specs and hardware requirements.<\/p>\n","protected":false},"author":1,"featured_media":4141,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[822,774,316,165,825,117],"class_list":["post-410","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-deepseek-ai-en","tag-deepseek-v4-1-flash-en","tag-hugging-face-en","tag-moe-en","tag-multimodal-en","tag--en"],"lang":"en","translations":{"en":410,"ja":408},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/410","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=410"}],"version-history":[{"count":72,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/410\/revisions"}],"predecessor-version":[{"id":14819,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/410\/revisions\/14819"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/4141"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=410"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=410"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=410"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}