--- title: NVIDIA DGX Spark performance · Ollama Blog | daily.dev description: Ollama published benchmark results for running local LLMs on the NVIDIA DGX Spark using release-day firmware (580.95.05) and Ollama v0.12.6. Tests measured... canonical: https://daily.dev/posts/nvidia-dgx-spark-performance-ollama-blog-cpfb88uh0 twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: NVIDIA DGX Spark performance · Ollama Blog | daily.dev og:description: Ollama published benchmark results for running local LLMs on the NVIDIA DGX Spark using release-day firmware (580.95.05) and Ollama v0.12.6. Tests measured... og:url: https://daily.dev/posts/nvidia-dgx-spark-performance-ollama-blog-cpfb88uh0 og:image: https://api.daily.dev/og/posts/cPfb88Uh0.png og:image:alt: NVIDIA DGX Spark performance · Ollama Blog og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # NVIDIA DGX Spark performance · Ollama Blog **[Ollama](https://daily.dev/sources/ollama)** · 3 min read · 0 upvotes · 0 comments ## Summary Ollama published benchmark results for running local LLMs on the NVIDIA DGX Spark using release-day firmware (580.95.05) and Ollama v0.12.6. Tests measured prefill and decode token throughput across models including gpt-oss (20B/120B), gemma3, llama3.1, deepseek-r1, and qwen3 at various quantization levels, with a fixed methodology (10 runs, temperature 0, 500-token output, caching disabled). The post also covers firmware update instructions via CLI and shows how to pair OpenAI's Codex CLI with Ollama's gpt-oss models locally, including running the 120B model that fits entirely in the DGX Spark's 120GB unified VRAM. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What decode speed can I expect running gpt-oss:120b on an NVIDIA DGX Spark with Ollama? The gpt-oss 120B model (MXFP4 quantization) achieves about 41.14 tokens per second decode speed and 1.169k tokens per second prefill on an NVIDIA DGX Spark, running Ollama v0.12.6 with firmware 580.95.05. The model fits entirely within the 120GB of unified VRAM provided by the GB10 Grace Blackwell Superchip, avoiding offloading. _Track real-world inference benchmarks like this on daily.dev before picking hardware for local LLM workloads._ ### How do I update the firmware on an NVIDIA DGX Spark from the command line? Update both the Ubuntu distribution and firmware together using apt and fwupdmgr: run sudo apt update, sudo apt dist-upgrade, sudo fwupdmgr refresh, sudo fwupdmgr upgrade, then sudo reboot. Alternatively, the DGX Dashboard can perform the update for firmware versions below 580.95.05. _Developers keeping DGX Spark firmware current follow release notes like this on daily.dev._ ### How do I run OpenAI's Codex CLI with a local gpt-oss model through Ollama? Install Codex with npm install -g @openai/codex, then run it with codex --oss --model gpt-oss to use the local model served by Ollama. For the larger variant, use codex --oss --model gpt-oss:120b, which runs entirely on-device given sufficient VRAM, such as the 120GB available on an NVIDIA DGX Spark. _Developers wiring local models into coding agents like Codex follow setup guides like this on daily.dev._ ## Similar posts on daily.dev - [dgx-lab-benchmarks-vs-reality-day-4](https://daily.dev/posts/dgx-lab-benchmarks-vs-reality-day-4-twf6imz0n) · Hacker News · 0 upvotes · 0 comments - [Run Local AI Agents with Faster Models and Multi-Node Clustering on NVIDIA DGX Spark](https://daily.dev/posts/run-local-ai-agents-with-faster-models-and-multi-node-clustering-on-nvidia-dgx-spark-moqgba8ql) · NVIDIA Developer · 1 upvotes · 0 comments - [NVIDIA DGX Spark: great hardware, early days for the ecosystem](https://daily.dev/posts/nvidia-dgx-spark-great-hardware-early-days-for-the-ecosystem-8avhsyx0d) · Simon Willison · 6 upvotes · 0 comments - [My 7-year-old GPU runs local AI perfectly, and I don't need my cloud subscriptions anymore](https://daily.dev/posts/my-7-year-old-gpu-runs-local-ai-perfectly-and-i-don-t-need-my-cloud-subscriptions-anymore-lvb6vhriu) · XDA Developers · 2 upvotes · 0 comments --- Tags: [#data-science](https://daily.dev/tags/data-science), [#ai-inference](https://daily.dev/tags/ai-inference), [#ollama](https://daily.dev/tags/ollama) [View this post on daily.dev](https://daily.dev/posts/nvidia-dgx-spark-performance-ollama-blog-cpfb88uh0) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"NVIDIA DGX Spark performance · Ollama Blog","url":"https://daily.dev/posts/nvidia-dgx-spark-performance-ollama-blog-cpfb88uh0","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/nvidia-dgx-spark-performance-ollama-blog-cpfb88uh0"},"datePublished":"2026-08-23T12:20:45.635Z","dateModified":"2026-08-23T12:42:15.904Z","description":"Ollama published benchmark results for running local LLMs on the NVIDIA DGX Spark using release-day firmware (580.95.05) and Ollama v0.12.6. Tests measured...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/03fb1a10ef158f8986c9778334c2af42?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/03fb1a10ef158f8986c9778334c2af42?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Ollama","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Ollama","logo":"https://media.daily.dev/image/upload/s--dRYYkBCj--/f_auto,q_auto/v1787487632/logos/ollama-blog?_a=BAMAMicg0","url":"https://daily.dev/sources/ollama"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/nvidia-dgx-spark-performance-ollama-blog-cpfb88uh0","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"data-science,ai-inference,ollama","timeRequired":"PT3M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Ollama","item":"https://daily.dev/sources/ollama"},{"@type":"ListItem","position":3,"name":"NVIDIA DGX Spark performance · Ollama Blog"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/nvidia-dgx-spark-performance-ollama-blog-cpfb88uh0#faq","mainEntity":[{"@type":"Question","name":"What decode speed can I expect running gpt-oss:120b on an NVIDIA DGX Spark with Ollama?","acceptedAnswer":{"@type":"Answer","text":"The gpt-oss 120B model (MXFP4 quantization) achieves about 41.14 tokens per second decode speed and 1.169k tokens per second prefill on an NVIDIA DGX Spark, running Ollama v0.12.6 with firmware 580.95.05. The model fits entirely within the 120GB of unified VRAM provided by the GB10 Grace Blackwell Superchip, avoiding offloading. Track real-world inference benchmarks like this on daily.dev before picking hardware for local LLM workloads."}},{"@type":"Question","name":"How do I update the firmware on an NVIDIA DGX Spark from the command line?","acceptedAnswer":{"@type":"Answer","text":"Update both the Ubuntu distribution and firmware together using apt and fwupdmgr: run sudo apt update, sudo apt dist-upgrade, sudo fwupdmgr refresh, sudo fwupdmgr upgrade, then sudo reboot. Alternatively, the DGX Dashboard can perform the update for firmware versions below 580.95.05. Developers keeping DGX Spark firmware current follow release notes like this on daily.dev."}},{"@type":"Question","name":"How do I run OpenAI's Codex CLI with a local gpt-oss model through Ollama?","acceptedAnswer":{"@type":"Answer","text":"Install Codex with npm install -g @openai/codex, then run it with codex --oss --model gpt-oss to use the local model served by Ollama. For the larger variant, use codex --oss --model gpt-oss:120b, which runs entirely on-device given sufficient VRAM, such as the 120GB available on an NVIDIA DGX Spark. Developers wiring local models into coding agents like Codex follow setup guides like this on daily.dev."}}]} ```