--- title: Bring Agentic Coding On-Device with Foundry Local description: Foundry Local now runs the Qwen3.8-27B coding model on-device and connects it to GitHub Copilot. With 4-bit quantization the model shrinks to about 18 GB,... canonical: https://daily.dev/posts/bring-agentic-coding-on-device-with-foundry-local-v5peiycsu twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Bring Agentic Coding On-Device with Foundry Local | daily.dev og:description: Foundry Local now runs the Qwen3.8-27B coding model on-device and connects it to GitHub Copilot. With 4-bit quantization the model shrinks to about 18 GB,... og:url: https://daily.dev/posts/bring-agentic-coding-on-device-with-foundry-local-v5peiycsu og:image: https://api.daily.dev/og/posts/v5PeiyCsU.png og:image:alt: Bring Agentic Coding On-Device with Foundry Local 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. # Bring Agentic Coding On-Device with Foundry Local **[DevBlogs](https://daily.dev/sources/devblogs)** · 5 min read · 0 upvotes · 0 comments ## Summary Foundry Local now runs the Qwen3.8-27B coding model on-device and connects it to GitHub Copilot. With 4-bit quantization the model shrinks to about 18 GB, running on consumer GPUs with 24 GB or more of VRAM. Microsoft benchmarked it against llama.cpp on an RTX 5090 (32 GB VRAM) and an NVIDIA DGX Spark (128 GB unified memory), finding Foundry Local faster at every prompt length, with the gap widening past 131K tokens. Setup requires Foundry Local CLI 0.13.0+ and GitHub Copilot's Bring Your Own Key feature. Mac and AMD GPU support is planned next. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How much VRAM do I need to run Qwen3.8-27B locally with Foundry Local? You need a consumer GPU with at least 24 GB of VRAM. With 4-bit quantization, Qwen3.8-27B shrinks to approximately 18 GB, which fits within that memory budget. Foundry Local requires CLI version 0.13.0 or later, plus separately installed CUDA and compatible cuDNN runtimes, since Foundry Local does not bundle these dependencies. _daily.dev surfaces hardware requirements like these for developers weighing local versus cloud AI coding setups._ ### How does Foundry Local's performance compare to llama.cpp for local coding models? Foundry Local delivered higher generation speed than llama.cpp at every tested prompt length when running the 4-bit Qwen3.8-27B model, tested on an NVIDIA RTX 5090 (32 GB VRAM) and an NVIDIA DGX Spark (128 GB unified memory). The advantage widens beyond 131K tokens, which matters for repository-scale coding workflows that need sustained long-context performance. _developers choosing a local inference engine can track benchmarks like this on daily.dev before committing._ ### How do I connect a local model to GitHub Copilot instead of using cloud models? Use GitHub Copilot's Bring Your Own Key (BYOK) feature to point Copilot at a locally running Foundry Local endpoint. After installing Foundry Local CLI 0.13.0+, run 'foundry model load qwen3.8-27b', retrieve the API endpoint via 'foundry server status', then switch the model provider in the Copilot model picker to Foundry Local and select Qwen3.8-27B. _daily.dev helps developers keep up with BYOK and local-model setups as Copilot's provider options expand._ ## Similar posts on daily.dev - [Advancing the Quality-Efficiency Frontier for Coding AI on Windows PCs - Microsoft Foundry on Windows Blog](https://daily.dev/posts/advancing-the-quality-efficiency-frontier-for-coding-ai-on-windows-pcs---microsoft-foundry-on-window-6dtuzslm8) · DevBlogs · 0 upvotes · 0 comments - [Qwen3-Coder-Next Local Deployment: A Complete Developer Guide \(2026\)](https://daily.dev/posts/qwen3-coder-next-local-deployment-a-complete-developer-guide-2026--cbao4it4m) · SitePoint · 1 upvotes · 0 comments - [The Best Local Coding Models for Any Setup](https://daily.dev/posts/the-best-local-coding-models-for-any-setup-8cm3jivdh) · Kilo Blog · 1 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 --- Tags: [#github](https://daily.dev/tags/github), [#local-ai](https://daily.dev/tags/local-ai) [View this post on daily.dev](https://daily.dev/posts/bring-agentic-coding-on-device-with-foundry-local-v5peiycsu) ```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":"Bring Agentic Coding On-Device with Foundry Local","url":"https://daily.dev/posts/bring-agentic-coding-on-device-with-foundry-local-v5peiycsu","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/bring-agentic-coding-on-device-with-foundry-local-v5peiycsu"},"datePublished":"2026-11-05T14:04:37.071Z","dateModified":"2026-11-05T14:05:10.484Z","description":"Foundry Local now runs the Qwen3.8-27B coding model on-device and connects it to GitHub Copilot. With 4-bit quantization the model shrinks to about 18 GB,...","image":"https://media.daily.dev/image/upload/s--VDukGCjf--/f_auto/v1722860399/public/Placeholder%2002","thumbnailUrl":"https://media.daily.dev/image/upload/s--VDukGCjf--/f_auto/v1722860399/public/Placeholder%2002","isAccessibleForFree":true,"articleSection":"DevBlogs","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":"DevBlogs","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/8f68b453325f482ebeb73fb780092713","url":"https://daily.dev/sources/devblogs"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/bring-agentic-coding-on-device-with-foundry-local-v5peiycsu","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"github,local-ai","timeRequired":"PT5M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"DevBlogs","item":"https://daily.dev/sources/devblogs"},{"@type":"ListItem","position":3,"name":"Bring Agentic Coding On-Device with Foundry Local"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/bring-agentic-coding-on-device-with-foundry-local-v5peiycsu#faq","mainEntity":[{"@type":"Question","name":"How much VRAM do I need to run Qwen3.8-27B locally with Foundry Local?","acceptedAnswer":{"@type":"Answer","text":"You need a consumer GPU with at least 24 GB of VRAM. With 4-bit quantization, Qwen3.8-27B shrinks to approximately 18 GB, which fits within that memory budget. Foundry Local requires CLI version 0.13.0 or later, plus separately installed CUDA and compatible cuDNN runtimes, since Foundry Local does not bundle these dependencies. daily.dev surfaces hardware requirements like these for developers weighing local versus cloud AI coding setups."}},{"@type":"Question","name":"How does Foundry Local's performance compare to llama.cpp for local coding models?","acceptedAnswer":{"@type":"Answer","text":"Foundry Local delivered higher generation speed than llama.cpp at every tested prompt length when running the 4-bit Qwen3.8-27B model, tested on an NVIDIA RTX 5090 (32 GB VRAM) and an NVIDIA DGX Spark (128 GB unified memory). The advantage widens beyond 131K tokens, which matters for repository-scale coding workflows that need sustained long-context performance. developers choosing a local inference engine can track benchmarks like this on daily.dev before committing."}},{"@type":"Question","name":"How do I connect a local model to GitHub Copilot instead of using cloud models?","acceptedAnswer":{"@type":"Answer","text":"Use GitHub Copilot's Bring Your Own Key (BYOK) feature to point Copilot at a locally running Foundry Local endpoint. After installing Foundry Local CLI 0.13.0+, run 'foundry model load qwen3.8-27b', retrieve the API endpoint via 'foundry server status', then switch the model provider in the Copilot model picker to Foundry Local and select Qwen3.8-27B. daily.dev helps developers keep up with BYOK and local-model setups as Copilot's provider options expand."}}]} ```