GitHub radar
A voice-assistant AI that runs right on your computer
Nemotron 3.5 Bolt is a small model for fast local voice assistants and simple device interfaces. Once installed, you get a local assistant that answers without internet or subscriptions.
It's a small language model from NVIDIA built on top of Gemma. It answers by voice or text right on your device, with no cloud involved.
When it helps
Useful for a simple local voice assistant or embedded interface with no subscription. Not a fit for serious coding or complex reasoning — bigger models handle that.
Pros
- Runs on an ordinary laptop — from 8GB of RAM
- No subscription or cloud — everything runs on your machine
- Pick a file size that fits your computer's memory
- Understands dozens of languages, including Russian
Cons
- Built for simple tasks — not for complex code
- Won't replace a model for serious AI work
How to set it up — step by step
- 1Open your AI coding agent (Claude Code, Codex) inside your project folder.
- 2Send the agent the text from the block below.
- 3Let the agent pick the right version for your laptop's memory and download it from Hugging Face.
- 4Check that the agent launched the model through llama.cpp or Ollama.
- 5Ask the model a test question and confirm you get an answer.
Text for your agent
Copy this and send it to your agent — Claude Code, Codex, any of them:
Set up the NVIDIA Nemotron 3.5 Bolt E4B model locally: https://huggingface.co/unsloth/NVIDIA-Nemotron-3.5-Bolt-E4B-GGUF — pick the build that fits my laptop, run it through llama.cpp or Ollama, and show me how to talk to it.
A regular laptop with 8 GB of RAM or more — the lightest build takes about 3.5 GB, the balanced one about 5 GB.
Open on Hugging Face▌ More finds