--- title: These 4 self-hosted LLMs run surprisingly well on... description: A blogger shares personal experience running four self-hosted, quantized LLMs (Mistral 7B, Phi-4 Mini, Gemma-3 4B, Qwen 3.5 4B) locally via KoboldCpp on a... canonical: https://daily.dev/posts/these-4-self-hosted-llms-run-surprisingly-well-on-integrated-graphics-coh33ddr3 twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: These 4 self-hosted LLMs run surprisingly well on integrated graphics | daily.dev og:description: A blogger shares personal experience running four self-hosted, quantized LLMs (Mistral 7B, Phi-4 Mini, Gemma-3 4B, Qwen 3.5 4B) locally via KoboldCpp on a... og:url: https://daily.dev/posts/these-4-self-hosted-llms-run-surprisingly-well-on-integrated-graphics-coh33ddr3 og:image: https://api.daily.dev/og/posts/CoH33DDr3.png og:image:alt: These 4 self-hosted LLMs run surprisingly well on integrated graphics 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. # These 4 self-hosted LLMs run surprisingly well on integrated graphics **[XDA Developers](https://daily.dev/sources/xda-developers)** · 5 min read · 5 upvotes · 1 comments ## Summary A blogger shares personal experience running four self-hosted, quantized LLMs (Mistral 7B, Phi-4 Mini, Gemma-3 4B, Qwen 3.5 4B) locally via KoboldCpp on a laptop with only AMD integrated graphics (Ryzen 7 5000 series, 16GB RAM). Each model is matched to a use case: Mistral 7B for brainstorming, Phi-4 Mini for quick everyday tasks, Gemma-3 4B for lightweight reasoning, and Qwen 3.5 4B for coding help. The takeaway is that 3B-7B quantized models offer a practical speed/quality balance on modest hardware, while larger models become impractically slow without a dedicated GPU. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### Can integrated graphics run local LLMs well without a dedicated GPU? Yes, integrated graphics such as AMD Radeon Graphics paired with a Ryzen 7 5000 series CPU and 16GB RAM can run quantized models in the 3B to 7B parameter range at practical speeds using KoboldCpp. Models like Mistral 7B (Q8_0), Phi-4 Mini (Q6_K), Gemma 3 4B (Q4_K_M), and Qwen 3.5 4B (Q8_K_XL) performed well, while larger models became noticeably slower. _Anyone weighing a GPU purchase for local AI can compare integrated-graphics setups like this on daily.dev._ ### Which quantized model works best for local coding assistance on limited hardware? Qwen3.5-4B-UD-Q8_K_XL, an 8-bit quantized version of Qwen 3.5 4B, worked well as a local coding assistant on integrated graphics, handling code writing, error fixes, explanations, and snippet edits. Larger code generations took longer to finish on integrated graphics, but smaller coding tasks felt practical and the output rarely needed a full rewrite. _Developers picking a local coding model for modest hardware can track real-world comparisons like this on daily.dev._ ## Community discussion Top comments from developers on daily.dev. **@akkitto** · 1 upvotes > Electricity cost ## Similar posts on daily.dev - [I ran local AI models on a six-year-old laptop with no GPU, and they actually worked](https://daily.dev/posts/i-ran-local-ai-models-on-a-six-year-old-laptop-with-no-gpu-and-they-actually-worked-v9yuny0hp) · XDA Developers · 2 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 - [I tested every tiny local LLM worth running, and only three survived the cut](https://daily.dev/posts/i-tested-every-tiny-local-llm-worth-running-and-only-three-survived-the-cut-ldlh8vwq9) · XDA Developers · 0 upvotes · 0 comments - [Best Local LLM Models 2026](https://daily.dev/posts/best-local-llm-models-2026-e56ssvqkw) · SitePoint · 2 upvotes · 0 comments --- Tags: [#data-science](https://daily.dev/tags/data-science), [#local-ai](https://daily.dev/tags/local-ai) [View this post on daily.dev](https://daily.dev/posts/these-4-self-hosted-llms-run-surprisingly-well-on-integrated-graphics-coh33ddr3) ```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":"These 4 self-hosted LLMs run surprisingly well on integrated graphics","url":"https://daily.dev/posts/these-4-self-hosted-llms-run-surprisingly-well-on-integrated-graphics-coh33ddr3","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/these-4-self-hosted-llms-run-surprisingly-well-on-integrated-graphics-coh33ddr3"},"datePublished":"2026-08-16T21:32:13.355Z","dateModified":"2026-09-13T18:31:52.496Z","description":"A blogger shares personal experience running four self-hosted, quantized LLMs (Mistral 7B, Phi-4 Mini, Gemma-3 4B, Qwen 3.5 4B) locally via KoboldCpp on a...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/8e8cd57218edefe15a34403c5c4c2c56?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/8e8cd57218edefe15a34403c5c4c2c56?_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":1,"discussionUrl":"https://daily.dev/posts/these-4-self-hosted-llms-run-surprisingly-well-on-integrated-graphics-coh33ddr3","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":5},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":1}],"keywords":"data-science,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":"XDA Developers","item":"https://daily.dev/sources/xda-developers"},{"@type":"ListItem","position":3,"name":"These 4 self-hosted LLMs run surprisingly well on integrated graphics"}]} {"@context":"https://schema.org","@type":"WebPage","@id":"https://daily.dev/posts/these-4-self-hosted-llms-run-surprisingly-well-on-integrated-graphics-coh33ddr3","comment":[{"@type":"Comment","text":"Electricity cost","datePublished":"2026-08-17T06:24:38.423Z","url":"https://daily.dev/posts/CoH33DDr3#c-hZJxioQg4","author":{"@type":"Person","name":"Daniel","url":"https://daily.dev/akkitto","image":"https://media.daily.dev/image/upload/s--FtwJqX4c--/f_auto/v1754900041/avatars/avatar_29TCpY2hJR72V3BlxPXzX?_a=BAMClqZW0"},"interactionStatistic":{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1}}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/these-4-self-hosted-llms-run-surprisingly-well-on-integrated-graphics-coh33ddr3#faq","mainEntity":[{"@type":"Question","name":"Can integrated graphics run local LLMs well without a dedicated GPU?","acceptedAnswer":{"@type":"Answer","text":"Yes, integrated graphics such as AMD Radeon Graphics paired with a Ryzen 7 5000 series CPU and 16GB RAM can run quantized models in the 3B to 7B parameter range at practical speeds using KoboldCpp. Models like Mistral 7B (Q8_0), Phi-4 Mini (Q6_K), Gemma 3 4B (Q4_K_M), and Qwen 3.5 4B (Q8_K_XL) performed well, while larger models became noticeably slower. Anyone weighing a GPU purchase for local AI can compare integrated-graphics setups like this on daily.dev."}},{"@type":"Question","name":"Which quantized model works best for local coding assistance on limited hardware?","acceptedAnswer":{"@type":"Answer","text":"Qwen3.5-4B-UD-Q8_K_XL, an 8-bit quantized version of Qwen 3.5 4B, worked well as a local coding assistant on integrated graphics, handling code writing, error fixes, explanations, and snippet edits. Larger code generations took longer to finish on integrated graphics, but smaller coding tasks felt practical and the output rarely needed a full rewrite. Developers picking a local coding model for modest hardware can track real-world comparisons like this on daily.dev."}}]} ```