{"id":426,"date":"2026-09-10T21:18:04","date_gmt":"2026-09-10T12:18:04","guid":{"rendered":"https:\/\/localmodelwatch.tsuchitsuchi.com\/2026\/09\/10\/nvidia-releases-deepseek-v4-pro-nvfp4\/"},"modified":"2026-11-08T19:08:17","modified_gmt":"2026-11-08T10:08:17","slug":"nvidia-releases-deepseek-v4-pro-nvfp4","status":"publish","type":"post","link":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/09\/10\/nvidia-releases-deepseek-v4-pro-nvfp4\/","title":{"rendered":"DeepSeek-V4-Pro-0813-nvfp4-DSpark: ~1005GB 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\/nvidia\/DeepSeek-V4-Pro-0813-nvfp4-DSpark\">nvidia\/DeepSeek-V4-Pro-0813-nvfp4-DSpark<\/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-en\/\">DeepSeek-V4 guide (2 articles)<\/a><\/td>\n<\/tr>\n<tr>\n<td>Publisher guide<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/publisher-nvidia-en\/\">NVIDIA: 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>safetensors<\/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>NVIDIA has released &#8220;nvidia\/DeepSeek-V4-Pro-0813-nvfp4-DSpark&#8221;, a fully NVFP4-quantized version of DeepSeek AI&#8217;s MoE language model DeepSeek-V4-Pro-0813. This model is available for both commercial and non-commercial use, and operates as a single self-contained checkpoint by including the quantized DSpark draft head. The model was quantized using NVIDIA Model Optimizer and supports inference with SGLang and vLLM.<\/p>\n<h2>Specifications<\/h2>\n<ul>\n<li>Architecture: Transformer (Mixture of Experts [MoE] with hybrid attention)<\/li>\n<li>Parameter Count: 1.65T total parameters, 49B active parameters<\/li>\n<li>Context Length: Up to 1 million tokens<\/li>\n<\/ul>\n<h2>Performance<\/h2>\n<p>This model&#8217;s backbone weights are bit-for-bit identical to &#8220;nvidia\/DeepSeek-V4-Pro-0813-NVFP4&#8221;, with only the DSpark draft head differing. Consequently, the output results match the target model, and accuracy evaluation results carry over directly. Below is a comparison of pre- and post-quantization evaluation results from the model card.<\/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>Precision<\/th>\n<th>GPQA Diamond<\/th>\n<th>AA-LCR<\/th>\n<th>\u03c4\u00b2-Bench Telecom<\/th>\n<th>SciCode<\/th>\n<th>IFBench<\/th>\n<th>Terminal-Bench Hard<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>MXFP4 (source)<\/td>\n<td>88.51<\/td>\n<td>68.67<\/td>\n<td>96.49<\/td>\n<td>53.45<\/td>\n<td>76.53<\/td>\n<td>51.39<\/td>\n<\/tr>\n<tr>\n<td>NVFP4<\/td>\n<td>88.42<\/td>\n<td>69.33<\/td>\n<td>98.25<\/td>\n<td>53.75<\/td>\n<td>75.68<\/td>\n<td>50.69<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p>Additionally, evaluations by the publishers of the original model, &#8220;DeepSeek-V4-Pro-0813&#8221;, report scores of 42.7 \/ 60.0 on HLE (without\/with tools) and 87.9 on Terminal Bench 2.1.<\/p>\n<p>These benchmark results show that the model maintains high performance in advanced reasoning and agent tasks such as GPQA Diamond (measuring PhD-level difficulty in physics, chemistry, and biology) and Terminal-Bench (measuring capability as a terminal operation agent). The difference in scores between the MXFP4-format source version and the NVFP4 format is minimal, keeping accuracy degradation from quantization to a minimum. Furthermore, measurements on MT-Bench and SPEED-Bench demonstrating the effectiveness of DSpark speculative decoding show an overall average acceptance length of around 3.7 to 3.8, achieving efficient speedups via block drafting.<\/p>\n<h2>Strengths and Use Cases<\/h2>\n<p>The original model, DeepSeek-V4-Pro-0813, excels at advanced reasoning, agentic AI applications, tool-use scenarios, and complex problem-solving tasks such as mathematics, software engineering, and enterprise AI assistants. It features an optimized Transformer architecture employing hybrid attention (Compressed Sparse Attention and Heavily Compressed Attention) and Manifold-Constrained Hyper-Connections, supporting a context length of up to 1 million tokens and reasoning effort levels (<code>low<\/code>, <code>high<\/code>, <code>max<\/code>) for thinking mode control. It also integrates the DSpark speculative decoding module within the same checkpoint, enabling efficient text generation.<\/p>\n<p><!-- lmw:hardware --><\/p>\n<h2>Hardware Requirements<\/h2>\n<p><strong>Estimated requirements (calculated by Local Model Watch)<\/strong> \u2014 1650.5B parameters (taken from the base model deepseek-ai\/DeepSeek-V4-Pro-0813)<\/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 1005GB of VRAM (multi-GPU or CPU offload required)<\/td>\n<td>NVFP4<\/td>\n<td>837.4GB<\/td>\n<td>1004.9GB<\/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 2026-11-08): llama.cpp: registered, vLLM: registered, MLX (mlx-lm): not registered. &#8220;Not registered&#8221; means the name is absent from that registry today, not that the model cannot run.<\/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. 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\/nvidia\/DeepSeek-V4-Pro-0813-nvfp4-DSpark\">nvidia\/DeepSeek-V4-Pro-0813-nvfp4-DSpark<\/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<h2>How to Get It<\/h2>\n<p>This model is available on Hugging Face in the safetensors distribution format. It can be accessed and used from the Hugging Face repository page for both commercial and non-commercial purposes.<\/p>\n<p>To deploy using SGLang, run the following command:<\/p>\n<pre><code class=\"language-sh\">python3 -m sglang.launch_server \\\n    --model-path nvidia\/DeepSeek-V4-Pro-0813-nvfp4-DSpark \\\n    --tp 8 \\\n    --trust-remote-code \\\n    --tool-call-parser deepseekv4 \\\n    --reasoning-parser deepseek-v4\n<\/code><\/pre>\n<p>An example of deploying with vLLM is as follows:<\/p>\n<pre><code class=\"language-sh\">vllm serve nvidia\/DeepSeek-V4-Pro-0813-nvfp4-DSpark \\\n    --served-model-name nvidia\/DeepSeek-V4-Pro-0813-nvfp4-DSpark \\\n    --tensor-parallel-size 8 \\\n    --enable-expert-parallel \\\n    --reasoning-parser deepseek_v4 \\\n    --max-model-len 400000 \\\n    --gpu-memory-utilization 0.9 \\\n    --kv-cache-dtype fp8 \\\n    --max-num-batched-tokens 8192 \\\n    --enable-chunked-prefill \\\n    --model-loader-extra-config '{&quot;enable_multithread_load&quot;: true, &quot;num_threads&quot;: 128}' \\\n    --speculative-config '{&quot;method&quot;:&quot;dspark&quot;,&quot;num_speculative_tokens&quot;:7,&quot;draft_sample_method&quot;:&quot;greedy&quot;}'\n<\/code><\/pre>\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<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/09\/23\/ifm-k2-horizon-32b-nvfp4-released\/\">K2-Horizon-32B-NVFP4: Our Test Answers, 32GB+ VRAM<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/09\/23\/ifm-k2-horizon-375b-a23b-nvfp4-released-2\/\">K2-Horizon-375B-A23B-NVFP4 Text Generation Model: ~257GB Memory<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/10\/02\/nvidia-releases-pixelumm\/\">PixelUMM Multimodal 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 does not fit even in 80GB; its smallest build needs about 1005GB) \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-en\/\">DeepSeek-V4 family overview (2 articles, 3 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 NVFP4 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-nvidia-en\/\">NVIDIA: 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\/nvidia\/DeepSeek-V4-Pro-0813-nvfp4-DSpark\">https:\/\/huggingface.co\/nvidia\/DeepSeek-V4-Pro-0813-nvfp4-DSpark<\/a><\/li>\n<li><a href=\"https:\/\/huggingface.co\/deepseek-ai\/DeepSeek-V4-Pro-0813\">https:\/\/huggingface.co\/deepseek-ai\/DeepSeek-V4-Pro-0813<\/a><\/li>\n<li><a href=\"https:\/\/github.com\/NVIDIA\/Model-Optimizer\">https:\/\/github.com\/NVIDIA\/Model-Optimizer<\/a><\/li>\n<\/ul>\n<p><!-- lmw:updates --><\/p>\n<h2>Update History<\/h2>\n<ul>\n<li>2026-09-20: Rewrote the article from re-collected sources and restored it from draft to published.<\/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>NVIDIA has released nvidia\/DeepSeek-V4-Pro-0813-nvfp4-DSpark, a fully NVFP4-quantized MoE language model supporting SGLang and vLLM inference.<\/p>\n","protected":false},"author":1,"featured_media":3950,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[310],"tags":[811,165,813,695,169,1547,592],"class_list":["post-426","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-new-models","tag-deepseek-v4-pro-en","tag-moe-en","tag-nvfp4-en","tag-nvidia-en","tag-sglang-en","tag-verified","tag-vllm-en"],"lang":"en","translations":{"en":426,"ja":424},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/426","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=426"}],"version-history":[{"count":71,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/426\/revisions"}],"predecessor-version":[{"id":15523,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/426\/revisions\/15523"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/3950"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=426"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=426"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=426"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}