--- title: I ran a local LLM entirely off a NAS, and it turned my... description: Running local LLMs on a NAS (TerraMaster F4-424 Max with TrueNAS and an i5-1235U CPU) is viable for lightweight AI tasks. LocalAI was chosen as the inference... canonical: https://daily.dev/posts/i-ran-a-local-llm-entirely-off-a-nas-and-it-turned-my-storage-box-into-an-ai-hosting-workstation-79t9q5cfk twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: I ran a local LLM entirely off a NAS, and it turned my storage box into an AI-hosting workstation | daily.dev og:description: Running local LLMs on a NAS (TerraMaster F4-424 Max with TrueNAS and an i5-1235U CPU) is viable for lightweight AI tasks. LocalAI was chosen as the inference... og:url: https://daily.dev/posts/i-ran-a-local-llm-entirely-off-a-nas-and-it-turned-my-storage-box-into-an-ai-hosting-workstation-79t9q5cfk og:image: https://api.daily.dev/og/posts/79t9q5cfk.png og:image:alt: I ran a local LLM entirely off a NAS, and it turned my storage box into an AI-hosting workstation 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 ran a local LLM entirely off a NAS, and it turned my storage box into an AI-hosting workstation **[XDA Developers](https://daily.dev/sources/xda-developers)** · 3 min read · 0 upvotes · 0 comments ## Summary Running local LLMs on a NAS (TerraMaster F4-424 Max with TrueNAS and an i5-1235U CPU) is viable for lightweight AI tasks. LocalAI was chosen as the inference engine over llama.cpp due to easier deployment on TrueNAS. With 8GB RAM available, Gemma-4-E2B runs at 10.5 tokens/second for text and supports image detection via MMPROJ files. The nomic-embed-text-v1.5 embedding model is also deployed alongside it to enable RAG capabilities by converting data into vectors for a vector database. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What token generation speed can I expect running Gemma-4-E2B on an Intel i5-1235U CPU? Gemma-4-E2B delivers about 10.5 tokens per second for text prompts and simple document inference tasks on an i5-1235U CPU. This is roughly double the performance of a Raspberry Pi. The model plus LocalAI together consume approximately 6.2GB of RAM when actively prompted, making it feasible on systems with 8GB allocated to AI workloads. _Developers running edge LLM experiments on low-power hardware share benchmarks like these on daily.dev._ ### Why use LocalAI instead of llama.cpp on TrueNAS? LocalAI is easier to deploy on TrueNAS because it is available as an app template, whereas llama.cpp requires manual setup as a custom Docker container or LXC. The Intel variants of llama.cpp also failed to detect the processor during container startup. LocalAI adds a web UI, supports agent creation, MCP servers, vision MMPROJ files, and automatically detects which models fit available system resources. _Those choosing between inference engines for self-hosted NAS setups find practical comparisons like this on daily.dev._ ## Similar posts on daily.dev - [Gemma 4 runs fine on my laptop, but putting it on my NAS is what changed how I use it](https://daily.dev/posts/gemma-4-runs-fine-on-my-laptop-but-putting-it-on-my-nas-is-what-changed-how-i-use-it-kfb8q9ppv) · XDA Developers · 0 upvotes · 0 comments - [Running local models is good now](https://daily.dev/posts/running-local-models-is-good-now-q20vqhe78) · Lobsters · 1 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: [#rag](https://daily.dev/tags/rag), [#self-hosting](https://daily.dev/tags/self-hosting), [#local-ai](https://daily.dev/tags/local-ai) [View this post on daily.dev](https://daily.dev/posts/i-ran-a-local-llm-entirely-off-a-nas-and-it-turned-my-storage-box-into-an-ai-hosting-workstation-79t9q5cfk) ```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 ran a local LLM entirely off a NAS, and it turned my storage box into an AI-hosting workstation","url":"https://daily.dev/posts/i-ran-a-local-llm-entirely-off-a-nas-and-it-turned-my-storage-box-into-an-ai-hosting-workstation-79t9q5cfk","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/i-ran-a-local-llm-entirely-off-a-nas-and-it-turned-my-storage-box-into-an-ai-hosting-workstation-79t9q5cfk"},"datePublished":"2026-08-07T12:04:22.555Z","dateModified":"2026-08-08T04:49:12.355Z","description":"Running local LLMs on a NAS (TerraMaster F4-424 Max with TrueNAS and an i5-1235U CPU) is viable for lightweight AI tasks. LocalAI was chosen as the inference...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/89888d346806b44452235e9dc470316d?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/89888d346806b44452235e9dc470316d?_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-ran-a-local-llm-entirely-off-a-nas-and-it-turned-my-storage-box-into-an-ai-hosting-workstation-79t9q5cfk","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"rag,self-hosting,local-ai","timeRequired":"PT3M"} {"@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 ran a local LLM entirely off a NAS, and it turned my storage box into an AI-hosting workstation"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/i-ran-a-local-llm-entirely-off-a-nas-and-it-turned-my-storage-box-into-an-ai-hosting-workstation-79t9q5cfk#faq","mainEntity":[{"@type":"Question","name":"What token generation speed can I expect running Gemma-4-E2B on an Intel i5-1235U CPU?","acceptedAnswer":{"@type":"Answer","text":"Gemma-4-E2B delivers about 10.5 tokens per second for text prompts and simple document inference tasks on an i5-1235U CPU. This is roughly double the performance of a Raspberry Pi. The model plus LocalAI together consume approximately 6.2GB of RAM when actively prompted, making it feasible on systems with 8GB allocated to AI workloads. Developers running edge LLM experiments on low-power hardware share benchmarks like these on daily.dev."}},{"@type":"Question","name":"Why use LocalAI instead of llama.cpp on TrueNAS?","acceptedAnswer":{"@type":"Answer","text":"LocalAI is easier to deploy on TrueNAS because it is available as an app template, whereas llama.cpp requires manual setup as a custom Docker container or LXC. The Intel variants of llama.cpp also failed to detect the processor during container startup. LocalAI adds a web UI, supports agent creation, MCP servers, vision MMPROJ files, and automatically detects which models fit available system resources. Those choosing between inference engines for self-hosted NAS setups find practical comparisons like this on daily.dev."}}]} ```