--- title: Coding From the Himalayas With an Offline AI Agent description: A Thomson Reuters engineer ran a coding agent entirely offline by pairing OpenCode with Ollama serving a local Qwen3.5 9B model. After downloading the model... canonical: https://daily.dev/posts/coding-from-the-himalayas-with-an-offline-ai-agent-kxvmgyfwv twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Coding From the Himalayas With an Offline AI Agent | daily.dev og:description: A Thomson Reuters engineer ran a coding agent entirely offline by pairing OpenCode with Ollama serving a local Qwen3.5 9B model. After downloading the model... og:url: https://daily.dev/posts/coding-from-the-himalayas-with-an-offline-ai-agent-kxvmgyfwv og:image: https://api.daily.dev/og/posts/kxVmgyfWv.png og:image:alt: Coding From the Himalayas With an Offline AI Agent 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. # Coding From the Himalayas With an Offline AI Agent **[Latent Space](https://daily.dev/sources/latentspace)** · 11 min read · 0 upvotes · 0 comments ## Summary A Thomson Reuters engineer ran a coding agent entirely offline by pairing OpenCode with Ollama serving a local Qwen3.5 9B model. After downloading the model and project files while online, she disabled Wi-Fi and confirmed external requests failed while localhost still worked. The agent found a bug in an expense calculator, proposed a one-line fix changing a reassignment to an accumulation, and five unit tests passed. Voltti budgets 12 to 16 GB of memory for the 6.6 GB Qwen3.5 9B build, and the author notes her 64k-token configuration may need more. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How do I run OpenCode with a local Ollama model completely offline? Configure opencode.json so both model and small_model point to Ollama's local endpoint (http://127.0.0.1:11434/v1), with only the local provider enabled. Start Ollama with OLLAMA_NO_CLOUD=1 and pull the model (e.g. qwen3.5:9b) while online, then disconnect Wi-Fi and Ethernet; localhost requests still work since they stay inside the machine, while external requests fail. _Developers testing offline AI workflows can track setup guides like this one on daily.dev._ ### How much RAM does Qwen3.5 9B need to run locally with Ollama? The Qwen3.5 9B build is about 6.6 GB to download, and Voltti recommends budgeting 12 to 16 GB of memory to run it. A larger context window, such as a 64k-token OpenCode configuration, may need more memory, so extra room should be left for context and other running apps. _Anyone sizing hardware for local LLM coding agents can compare setups like this on daily.dev._ ### How does OpenCode compare to Goose and OpenHands-SDK in coding agent benchmarks? On 50 Terminal-Bench Pro tasks using cloud models, Goose scored 48% and 38% pass rates across two models with 95% confidence intervals, OpenHands-SDK scored 50% and 46%, per the preprint 'The Scaffold Effect in Coding Agents.' The authors caution most differences were not statistically clear, but OpenCode used roughly 40 times more tokens per solved task than Goose. _Developers choosing between coding agent scaffolds can follow benchmark comparisons like this on daily.dev._ ## Similar posts on daily.dev - [Local AI Models for Coding 2026: Ollama Setup and Best Models](https://daily.dev/posts/local-ai-models-for-coding-2026-ollama-setup-and-best-models-sohghmj3p) · Alex CloudStar · 1 upvotes · 0 comments - [I think we're cooked...](https://daily.dev/posts/i-think-we-re-cooked--c98iytsiv) · DEV · 47 upvotes · 16 comments - [Run Coding Agents on Local AI — Zero Cloud, Full Control](https://daily.dev/posts/run-coding-agents-on-local-ai-zero-cloud-full-control-i0ym3cjcg) · ITNEXT · 0 upvotes · 0 comments --- Tags: [#local-ai](https://daily.dev/tags/local-ai), [#ollama](https://daily.dev/tags/ollama), [#qwen](https://daily.dev/tags/qwen), [#opencode](https://daily.dev/tags/opencode) [View this post on daily.dev](https://daily.dev/posts/coding-from-the-himalayas-with-an-offline-ai-agent-kxvmgyfwv) ```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":"Coding From the Himalayas With an Offline AI Agent","url":"https://daily.dev/posts/coding-from-the-himalayas-with-an-offline-ai-agent-kxvmgyfwv","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/coding-from-the-himalayas-with-an-offline-ai-agent-kxvmgyfwv"},"datePublished":"2026-10-27T14:50:42.574Z","dateModified":"2026-10-27T14:52:59.500Z","description":"A Thomson Reuters engineer ran a coding agent entirely offline by pairing OpenCode with Ollama serving a local Qwen3.5 9B model. After downloading the model...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/cbf6d18b4d02d8f2dbc4cb7a16e8f564?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/cbf6d18b4d02d8f2dbc4cb7a16e8f564?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Latent Space","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":"Latent Space","logo":"https://media.daily.dev/image/upload/s--DnJ9laFj--/f_auto,q_auto/v1780213196/logos/latentspace?_a=BAMAMiWQ0","url":"https://daily.dev/sources/latentspace"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/coding-from-the-himalayas-with-an-offline-ai-agent-kxvmgyfwv","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"local-ai,ollama,qwen,opencode","timeRequired":"PT11M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Latent Space","item":"https://daily.dev/sources/latentspace"},{"@type":"ListItem","position":3,"name":"Coding From the Himalayas With an Offline AI Agent"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/coding-from-the-himalayas-with-an-offline-ai-agent-kxvmgyfwv#faq","mainEntity":[{"@type":"Question","name":"How do I run OpenCode with a local Ollama model completely offline?","acceptedAnswer":{"@type":"Answer","text":"Configure opencode.json so both model and small_model point to Ollama's local endpoint (http://127.0.0.1:11434/v1), with only the local provider enabled. Start Ollama with OLLAMA_NO_CLOUD=1 and pull the model (e.g. qwen3.5:9b) while online, then disconnect Wi-Fi and Ethernet; localhost requests still work since they stay inside the machine, while external requests fail. Developers testing offline AI workflows can track setup guides like this one on daily.dev."}},{"@type":"Question","name":"How much RAM does Qwen3.5 9B need to run locally with Ollama?","acceptedAnswer":{"@type":"Answer","text":"The Qwen3.5 9B build is about 6.6 GB to download, and Voltti recommends budgeting 12 to 16 GB of memory to run it. A larger context window, such as a 64k-token OpenCode configuration, may need more memory, so extra room should be left for context and other running apps. Anyone sizing hardware for local LLM coding agents can compare setups like this on daily.dev."}},{"@type":"Question","name":"How does OpenCode compare to Goose and OpenHands-SDK in coding agent benchmarks?","acceptedAnswer":{"@type":"Answer","text":"On 50 Terminal-Bench Pro tasks using cloud models, Goose scored 48% and 38% pass rates across two models with 95% confidence intervals, OpenHands-SDK scored 50% and 46%, per the preprint 'The Scaffold Effect in Coding Agents.' The authors caution most differences were not statistically clear, but OpenCode used roughly 40 times more tokens per solved task than Goose. Developers choosing between coding agent scaffolds can follow benchmark comparisons like this on daily.dev."}}]} ```