--- title: Can a Local LLM Run My AI Assistant? | daily.dev description: A developer replayed 27 real production tasks from a 90-tool personal AI agent (Jarvis) against two local models, benchmarking them against a frozen Claude... canonical: https://daily.dev/posts/can-a-local-llm-run-my-ai-assistant--k8pjpamza twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Can a Local LLM Run My AI Assistant? | daily.dev og:description: A developer replayed 27 real production tasks from a 90-tool personal AI agent (Jarvis) against two local models, benchmarking them against a frozen Claude... og:url: https://daily.dev/posts/can-a-local-llm-run-my-ai-assistant--k8pjpamza og:image: https://api.daily.dev/og/posts/k8PJPaMzA.png og:image:alt: Can a Local LLM Run My AI Assistant? 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. # Can a Local LLM Run My AI Assistant? **[Towards Data Science](https://daily.dev/sources/tds)** · 15 min read · 0 upvotes · 0 comments ## Summary A developer replayed 27 real production tasks from a 90-tool personal AI agent (Jarvis) against two local models, benchmarking them against a frozen Claude baseline. On a single RTX 3090 with a 16K context window, qwen3-coder:30B scored 22.8/100 vs Claude's 89.4, with malformed tool-call syntax appearing in 25.9% of answers. After upgrading to three RTX 3090s (72 GB VRAM) and running a 122B MoE model at 256K context, the score jumped to 80.0/100 — 89.4% of Claude's performance — with zero malformed tool calls. The local model costs $0.000969 per task vs Claude's $0.763, roughly 787× cheaper. The author switched Jarvis to the local model and cancelled Claude Max, while noting the key caveat: model size and context window changed simultaneously, making it impossible to isolate which factor drove the improvement. The hypothesis is that context window size mattered more than parameter count, since the most characteristic failures in round 1 match what you'd expect from truncated tool schemas. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How does a local 122B LLM compare to Claude when running a multi-tool AI agent? A 122B MoE model at Q3_K_M on three RTX 3090s (72 GB VRAM) with a 256K context window scored 80.0/100 against Claude's 89.4 on 27 real production tasks across a 90-tool personal agent — 89.4% of Claude's quality at roughly 787× lower cost ($0.000969 vs $0.763 per task). The local model eliminated malformed tool-call syntax entirely and beat Claude outright on general tasks. _Developers choosing between local and API-hosted models for agent workloads track findings like these on daily.dev._ ### Why does a local LLM produce malformed tool calls in a LangGraph agent? Malformed tool calls — raw text like ... leaking into user-facing responses instead of proper LangGraph tool invocations — are strongly associated with a truncated context window. A 30B model on a 16K context window produced malformed calls in 7 of 27 tasks (25.9%), while the same agent with a 256K context window and a 122B model produced zero malformed calls. The hypothesis is that tool schemas were being cut off mid-definition at 16K. _Teams debugging agent tool-call failures in LangGraph find relevant production experiences on daily.dev._ ### What is the electricity cost per task when running a 122B LLM locally on three RTX 3090s? Running a 122B MoE model at Q3_K_M across three RTX 3090s costs approximately $0.000969 per task, based on 0.1573 kWh measured across 27 tasks at a home dual-rate electricity tariff. This is 787× cheaper than Claude's $0.763 per task API cost. The three cards draw about 107 W at idle just keeping the model resident, so per-task cost is higher if the model stays loaded permanently versus spinning up on demand. _Engineers pricing local GPU inference versus API costs for agent workloads share real numbers on daily.dev._ ## Similar posts on daily.dev - [Instead of paying $40 a month for cloud AI tools, I run everything through a local model on my gaming PC](https://daily.dev/posts/instead-of-paying-40-a-month-for-cloud-ai-tools-i-run-everything-through-a-local-model-on-my-gamin-gyzicbsx7) · XDA Developers · 4 upvotes · 1 comments - [\[Revised\] You Don’t Need to Spend $100/mo on Claude Code: Your Guide to Local Coding Models](https://daily.dev/posts/revised-you-don-t-need-to-spend-100-mo-on-claude-code-your-guide-to-local-coding-models-9orgfaed1) · Hacker News · 2 upvotes · 0 comments - [I stopped my local LLM agent from running out of context, and now it handles what Claude Code does](https://daily.dev/posts/i-stopped-my-local-llm-agent-from-running-out-of-context-and-now-it-handles-what-claude-code-does-kfav6uv6k) · XDA Developers · 0 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#claude](https://daily.dev/tags/claude), [#local-ai](https://daily.dev/tags/local-ai), [#langgraph](https://daily.dev/tags/langgraph), [#qwen](https://daily.dev/tags/qwen) [View this post on daily.dev](https://daily.dev/posts/can-a-local-llm-run-my-ai-assistant--k8pjpamza) ```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":"Can a Local LLM Run My AI Assistant?","url":"https://daily.dev/posts/can-a-local-llm-run-my-ai-assistant--k8pjpamza","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/can-a-local-llm-run-my-ai-assistant--k8pjpamza"},"datePublished":"2026-08-11T12:38:39.863Z","dateModified":"2026-08-11T12:39:17.177Z","description":"A developer replayed 27 real production tasks from a 90-tool personal AI agent (Jarvis) against two local models, benchmarking them against a frozen Claude...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/62826e57178a8fdc0d92cf97c27b81c8?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/62826e57178a8fdc0d92cf97c27b81c8?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Towards Data Science","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":"Towards Data Science","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/tds","url":"https://daily.dev/sources/tds"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/can-a-local-llm-run-my-ai-assistant--k8pjpamza","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,claude,local-ai,langgraph,qwen","timeRequired":"PT15M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Towards Data Science","item":"https://daily.dev/sources/tds"},{"@type":"ListItem","position":3,"name":"Can a Local LLM Run My AI Assistant?"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/can-a-local-llm-run-my-ai-assistant--k8pjpamza#faq","mainEntity":[{"@type":"Question","name":"How does a local 122B LLM compare to Claude when running a multi-tool AI agent?","acceptedAnswer":{"@type":"Answer","text":"A 122B MoE model at Q3_K_M on three RTX 3090s (72 GB VRAM) with a 256K context window scored 80.0/100 against Claude's 89.4 on 27 real production tasks across a 90-tool personal agent — 89.4% of Claude's quality at roughly 787× lower cost ($0.000969 vs $0.763 per task). The local model eliminated malformed tool-call syntax entirely and beat Claude outright on general tasks. Developers choosing between local and API-hosted models for agent workloads track findings like these on daily.dev."}},{"@type":"Question","name":"Why does a local LLM produce malformed tool calls in a LangGraph agent?","acceptedAnswer":{"@type":"Answer","text":"Malformed tool calls — raw text like ... leaking into user-facing responses instead of proper LangGraph tool invocations — are strongly associated with a truncated context window. A 30B model on a 16K context window produced malformed calls in 7 of 27 tasks (25.9%), while the same agent with a 256K context window and a 122B model produced zero malformed calls. The hypothesis is that tool schemas were being cut off mid-definition at 16K. Teams debugging agent tool-call failures in LangGraph find relevant production experiences on daily.dev."}},{"@type":"Question","name":"What is the electricity cost per task when running a 122B LLM locally on three RTX 3090s?","acceptedAnswer":{"@type":"Answer","text":"Running a 122B MoE model at Q3_K_M across three RTX 3090s costs approximately $0.000969 per task, based on 0.1573 kWh measured across 27 tasks at a home dual-rate electricity tariff. This is 787× cheaper than Claude's $0.763 per task API cost. The three cards draw about 107 W at idle just keeping the model resident, so per-task cost is higher if the model stays loaded permanently versus spinning up on demand. Engineers pricing local GPU inference versus API costs for agent workloads share real numbers on daily.dev."}}]} ```