--- title: Local AI Coding Agents Are Finally Good Enough | daily.dev description: Qwen 3.6, a 27B parameter open-source model optimized for agentic coding, now runs locally on Apple Silicon machines with enough RAM (24GB+ recommended). Using... canonical: https://daily.dev/posts/local-ai-coding-agents-are-finally-good-enough-1zh1f2ttq twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Local AI Coding Agents Are Finally Good Enough | daily.dev og:description: Qwen 3.6, a 27B parameter open-source model optimized for agentic coding, now runs locally on Apple Silicon machines with enough RAM (24GB+ recommended). Using... og:url: https://daily.dev/posts/local-ai-coding-agents-are-finally-good-enough-1zh1f2ttq og:image: https://api.daily.dev/og/posts/1zH1f2ttq.png og:image:alt: Local AI Coding Agents Are Finally Good Enough 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. # Local AI Coding Agents Are Finally Good Enough **[Code with Beto](https://daily.dev/sources/codewithbeto)** · 16 min read · 3 upvotes · 0 comments ## Summary Qwen 3.6, a 27B parameter open-source model optimized for agentic coding, now runs locally on Apple Silicon machines with enough RAM (24GB+ recommended). Using LM Studio for model management and OpenCode as the coding agent interface, the model can read codebases, make multi-step tool calls, recognize images, and generate working code — all offline. Benchmarks show it performs comparably to Claude 4.5 on agentic coding tasks. The MLX framework from Apple significantly improves inference speed and reliability. Practical demos include building a Minesweeper game, a Snake game, and modifying a real mobile app codebase. Hardware requirements are steep — an M3 Ultra with 96GB was used — and an 18GB MacBook Pro couldn't run it. LM Studio also supports remote access via phone through its LM Link feature. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What are the hardware requirements to run Qwen 3.6 27B locally on a Mac? At least 16 GB of memory is listed as the minimum, but 24 GB or more is recommended since 16 GB proves tight in practice. A MacBook Pro with 18 GB of VRAM could not run the model, while an Apple M3 Ultra with 96 GB of VRAM and 80 cores ran it smoothly, including complex agentic coding tasks that took around ten minutes. _daily.dev helps developers compare real-world hardware requirements before committing to local AI coding setups._ ### How does Qwen 3.6 27B compare to Claude Opus for agentic coding? It performs close to Claude Opus 4.5 on agentic terminal benchmarks, though it falls short of the newer Opus 4.8. It surpasses its own predecessor, Qwen 3.5, on agentic coding benchmarks and is fully open source, making it viable for offline coding tasks like refactoring and simple feature additions when a cloud subscription isn't available. _developers weighing local versus cloud coding agents can track these tradeoffs on daily.dev._ ### How do I connect a local model in LM Studio to Open Code for agentic coding? Enable developer mode in LM Studio, load the model, then turn on the local server endpoint, which is OpenAI API compatible. In Open Code's configuration, point to that local endpoint and add the modalities field to the JSON config to enable image recognition, then the local model becomes selectable within Open Code's local models list. _daily.dev keeps developers current on connecting local models to agentic coding tools like open code._ ## Similar posts on daily.dev - [Viability of local models for coding](https://daily.dev/posts/viability-of-local-models-for-coding-1cgbfrvir) · Martin Fowler · 0 upvotes · 0 comments - [Running local models on an M4 with 24GB memory](https://daily.dev/posts/running-local-models-on-an-m4-with-24gb-memory-iezpeho6j) · Hacker News · 4 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#lm-studio](https://daily.dev/tags/lm-studio), [#local-ai](https://daily.dev/tags/local-ai), [#mlx](https://daily.dev/tags/mlx), [#qwen](https://daily.dev/tags/qwen) [View this post on daily.dev](https://daily.dev/posts/local-ai-coding-agents-are-finally-good-enough-1zh1f2ttq) ```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":"Local AI Coding Agents Are Finally Good Enough","url":"https://daily.dev/posts/local-ai-coding-agents-are-finally-good-enough-1zh1f2ttq","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/local-ai-coding-agents-are-finally-good-enough-1zh1f2ttq"},"datePublished":"2026-07-09T10:26:41.533Z","dateModified":"2026-10-15T02:22:56.537Z","description":"Qwen 3.6, a 27B parameter open-source model optimized for agentic coding, now runs locally on Apple Silicon machines with enough RAM (24GB+ recommended). Using...","image":"https://i.ytimg.com/vi/Zof2Oaj14rk/sddefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/Zof2Oaj14rk/sddefault.jpg","isAccessibleForFree":true,"articleSection":"Code with Beto","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":"Code with Beto","logo":"https://media.daily.dev/image/upload/s--8oKUFd2X--/f_auto,q_auto/v1774179046/logos/codewithbeto?_a=BAMAMiWQ0","url":"https://daily.dev/sources/codewithbeto"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/local-ai-coding-agents-are-finally-good-enough-1zh1f2ttq","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":3},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,lm-studio,local-ai,mlx,qwen","timeRequired":"PT16M","video":{"@type":"VideoObject","name":"Local AI Coding Agents Are Finally Good Enough","description":"Qwen 3.6, a 27B parameter open-source model optimized for agentic coding, now runs locally on Apple Silicon machines with enough RAM (24GB+ recommended). Using...","thumbnailUrl":"https://i.ytimg.com/vi/Zof2Oaj14rk/sddefault.jpg","uploadDate":"2026-07-09T10:26:41.533Z","duration":"PT16M","url":"https://api.daily.dev/r/1zH1f2ttq","embedUrl":"https://www.youtube.com/embed/Zof2Oaj14rk"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Code with Beto","item":"https://daily.dev/sources/codewithbeto"},{"@type":"ListItem","position":3,"name":"Local AI Coding Agents Are Finally Good Enough"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/local-ai-coding-agents-are-finally-good-enough-1zh1f2ttq#faq","mainEntity":[{"@type":"Question","name":"What are the hardware requirements to run Qwen 3.6 27B locally on a Mac?","acceptedAnswer":{"@type":"Answer","text":"At least 16 GB of memory is listed as the minimum, but 24 GB or more is recommended since 16 GB proves tight in practice. A MacBook Pro with 18 GB of VRAM could not run the model, while an Apple M3 Ultra with 96 GB of VRAM and 80 cores ran it smoothly, including complex agentic coding tasks that took around ten minutes. daily.dev helps developers compare real-world hardware requirements before committing to local AI coding setups."}},{"@type":"Question","name":"How does Qwen 3.6 27B compare to Claude Opus for agentic coding?","acceptedAnswer":{"@type":"Answer","text":"It performs close to Claude Opus 4.5 on agentic terminal benchmarks, though it falls short of the newer Opus 4.8. It surpasses its own predecessor, Qwen 3.5, on agentic coding benchmarks and is fully open source, making it viable for offline coding tasks like refactoring and simple feature additions when a cloud subscription isn't available. developers weighing local versus cloud coding agents can track these tradeoffs on daily.dev."}},{"@type":"Question","name":"How do I connect a local model in LM Studio to Open Code for agentic coding?","acceptedAnswer":{"@type":"Answer","text":"Enable developer mode in LM Studio, load the model, then turn on the local server endpoint, which is OpenAI API compatible. In Open Code's configuration, point to that local endpoint and add the modalities field to the JSON config to enable image recognition, then the local model becomes selectable within Open Code's local models list. daily.dev keeps developers current on connecting local models to agentic coding tools like open code."}}]} ```