lab / model
Local AI
Chat with a small language model that runs on your own machine.
Runs on your device over WebGPU.
Local AI
Qwen2.5 0.5B Instruct · Apache-2.0
A small instruct model that runs in this browser tab over WebGPU — no server, no API key. Clicking below downloads ~290 MB of 4-bit (q4f16) weights once; your browser caches them so the next visit is instant.
enable JavaScript to run it
waking…
runs in this tab over WebGPU — close the tab and the conversation is gone.
How it works
- Model: Qwen2.5 0.5B Instruct (4-bit (q4f16)), Apache-2.0-licensed, about ~290 MB of weights downloaded on first wake and cached by your browser for next time.
- Runtime: WebGPU — needs a recent Chrome, Edge, or other WebGPU-capable browser.
- Knowledge: the model is told only the facts on this site. Ask it about the projects or the approach; ask it anything else and it will tell you that's not on the site.
- Privacy: no server call and no API key. The conversation lives in this tab and disappears when you close it.