--- title: I expected Claude Code to crush its open-source rival on... description: A local open-source coding agent matched Claude Code on a real task while costing far less. Running Qwen3.6-35B-A3B through OpenCode on a 16GB RTX 4070 Ti... canonical: https://daily.dev/posts/i-expected-claude-code-to-crush-its-open-source-rival-on-my-16gb-gpu-but-that-s-not-what-happened-hzmdfvvdf twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: I expected Claude Code to crush its open-source rival on my 16GB GPU, but that’s not what happened | daily.dev og:description: A local open-source coding agent matched Claude Code on a real task while costing far less. Running Qwen3.6-35B-A3B through OpenCode on a 16GB RTX 4070 Ti... og:url: https://daily.dev/posts/i-expected-claude-code-to-crush-its-open-source-rival-on-my-16gb-gpu-but-that-s-not-what-happened-hzmdfvvdf og:image: https://api.daily.dev/og/posts/hzMdfvVdf.png og:image:alt: I expected Claude Code to crush its open-source rival on my 16GB GPU, but that’s not what happened 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. # I expected Claude Code to crush its open-source rival on my 16GB GPU, but that’s not what happened **[XDA Developers](https://daily.dev/sources/xda-developers)** · 4 min read · 2 upvotes · 0 comments ## Summary A local open-source coding agent matched Claude Code on a real task while costing far less. Running Qwen3.6-35B-A3B through OpenCode on a 16GB RTX 4070 Ti Super, the author built a GPU logger that passed all six requirements in 3 minutes 44 seconds, fully offline. Claude Sonnet 5.5 finished the same job in 2 minutes 42 seconds with better chart polish, but the session cost $0.37 versus roughly a fifth of a cent in electricity for the local run. The author calls the result a tie on utility and questions the value of a $20 Claude subscription for workflows that don't need flagship capability. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### Can a 17.7GB Qwen3.6-35B-A3B model run on a 16GB GPU? Yes, because it is a mixture-of-experts model activating only around 3 billion parameters per token, allowing optimizations like llama.cpp's --n-cpu-moe flag to offload 12 layers of experts to system RAM. This setup leaves room for a 128K context window while using about 15GB of VRAM on an RTX 4070 Ti Super. _Developers eyeing local coding agents can track real GPU memory tricks like this on daily.dev._ ### How does Claude Code compare to a local Qwen3.6 agent on coding tasks? On a GPU logging task with clear pass/fail requirements, Claude Sonnet 5.5 via Claude Code finished in 2 minutes 42 seconds and Qwen3.6-35B-A3B via OpenCode finished in 3 minutes 44 seconds, both passing all six requirements without human intervention. Claude's session cost about $0.37 in subscription usage, while the local run cost roughly a fifth of a cent in electricity. _Anyone weighing a paid coding agent against a local setup can compare benchmarks like this on daily.dev._ ## Similar posts on daily.dev - [I run a 24GB GPU instead of paying for Claude or Codex, and Qwen 3.6 keeps up more than I expected](https://daily.dev/posts/i-run-a-24gb-gpu-instead-of-paying-for-claude-or-codex-and-qwen-3-6-keeps-up-more-than-i-expected-t74m2hhbi) · XDA Developers · 0 upvotes · 0 comments - [I split my coding work between Claude, Qwen3-Coder and Gemma 4, and it costs less than paying for one subscription](https://daily.dev/posts/i-split-my-coding-work-between-claude-qwen3-coder-and-gemma-4-and-it-costs-less-than-paying-for-on-r1us9xfzm) · XDA Developers · 0 upvotes · 0 comments - [Claude Code Sends 4.7x More Tokens Than OpenCode Before Reading Your Prompt](https://daily.dev/posts/claude-code-sends-4-7x-more-tokens-than-opencode-before-reading-your-prompt-9m02iom1z) · Hacker News · 6 upvotes · 0 comments - [Untitled](https://daily.dev/posts/untitled-9dyldngag) · SitePoint · 0 upvotes · 0 comments --- Tags: [#claude-code](https://daily.dev/tags/claude-code), [#local-ai](https://daily.dev/tags/local-ai), [#qwen](https://daily.dev/tags/qwen), [#opencode](https://daily.dev/tags/opencode), [#llama-cpp](https://daily.dev/tags/llama-cpp) [View this post on daily.dev](https://daily.dev/posts/i-expected-claude-code-to-crush-its-open-source-rival-on-my-16gb-gpu-but-that-s-not-what-happened-hzmdfvvdf) ```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":"I expected Claude Code to crush its open-source rival on my 16GB GPU, but that’s not what happened","url":"https://daily.dev/posts/i-expected-claude-code-to-crush-its-open-source-rival-on-my-16gb-gpu-but-that-s-not-what-happened-hzmdfvvdf","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/i-expected-claude-code-to-crush-its-open-source-rival-on-my-16gb-gpu-but-that-s-not-what-happened-hzmdfvvdf"},"datePublished":"2026-11-08T17:33:02.171Z","dateModified":"2026-11-08T17:35:56.589Z","description":"A local open-source coding agent matched Claude Code on a real task while costing far less. Running Qwen3.6-35B-A3B through OpenCode on a 16GB RTX 4070 Ti...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/da0f7d6ed99e0984672bbe5cf1cdcc20?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/da0f7d6ed99e0984672bbe5cf1cdcc20?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"XDA Developers","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":"XDA Developers","logo":"https://media.daily.dev/image/upload/s--kCg6yyAP--/f_auto,q_auto/v1774964407/logos/xda-developers?_a=BAMAMiWQ0","url":"https://daily.dev/sources/xda-developers"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/i-expected-claude-code-to-crush-its-open-source-rival-on-my-16gb-gpu-but-that-s-not-what-happened-hzmdfvvdf","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":2},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"claude-code,local-ai,qwen,opencode,llama-cpp","timeRequired":"PT4M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"XDA Developers","item":"https://daily.dev/sources/xda-developers"},{"@type":"ListItem","position":3,"name":"I expected Claude Code to crush its open-source rival on my 16GB GPU, but that’s not what happened"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/i-expected-claude-code-to-crush-its-open-source-rival-on-my-16gb-gpu-but-that-s-not-what-happened-hzmdfvvdf#faq","mainEntity":[{"@type":"Question","name":"Can a 17.7GB Qwen3.6-35B-A3B model run on a 16GB GPU?","acceptedAnswer":{"@type":"Answer","text":"Yes, because it is a mixture-of-experts model activating only around 3 billion parameters per token, allowing optimizations like llama.cpp's --n-cpu-moe flag to offload 12 layers of experts to system RAM. This setup leaves room for a 128K context window while using about 15GB of VRAM on an RTX 4070 Ti Super. Developers eyeing local coding agents can track real GPU memory tricks like this on daily.dev."}},{"@type":"Question","name":"How does Claude Code compare to a local Qwen3.6 agent on coding tasks?","acceptedAnswer":{"@type":"Answer","text":"On a GPU logging task with clear pass/fail requirements, Claude Sonnet 5.5 via Claude Code finished in 2 minutes 42 seconds and Qwen3.6-35B-A3B via OpenCode finished in 3 minutes 44 seconds, both passing all six requirements without human intervention. Claude's session cost about $0.37 in subscription usage, while the local run cost roughly a fifth of a cent in electricity. Anyone weighing a paid coding agent against a local setup can compare benchmarks like this on daily.dev."}}]} ```