# MiniPC to run a LLM w/ voice assistant - Best small LLM? — retained record

> Retained reddit record from reddit:LocalLLaMA.

Canonical: https://highsignal.app/data/records/23f85e931a36067f

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:LocalLLaMA)  published 2026-10-03 05:04 ingested 2026-10-04 02:38

[original source ↗](https://www.reddit.com/r/LocalLLaMA/comments/1wweqcf/minipc_to_run_a_llm_w_voice_assistant_best_small/)

## retained text



```
I'm having a go at building a fully offline, voice-first assistant running locally on a fanless mini PC (4 core Celeron J6412 w/ 16GB DDR4, 512GB SATA SSD, crappy Intel UHD iGPU only)  Going to try Ubuntu 24.04, llama.cpp, Python.

Are there any models that might be able to hold a strong persona and stay concise on this class of CPU?  Is there a STT for short commands that might work real time on a weak CPU?
```



document fetched 2026-10-04 00:36 · canonical https://reddit.com/r/LocalLLaMA/comments/1wweqcf/minipc_to_run_a_llm_w_voice_assistant_best_small
