Browser LLM

SmolLM2 · Runs on your CPU · No server · No API key

⚠ Runs entirely in your browser

This is SmolLM2-135M — a 135 million parameter model that fits in a browser tab. For context, GPT-4 has ~1.8 trillion. Expect short memory, simple reasoning, and occasional nonsense. First load downloads ~270 MB. Do not compare this to ChatGPT or any cloud model.

Downloading model — first load only, then cached…
System
Loading SmolLM2-135M-Instruct into your browser. This is a ~90MB download on first run, then cached locally. Inference runs 100% on your CPU via WebAssembly.
SmolLM2-135M-Instruct · Transformers.js v3 · WebAssembly · GitHub Pages compatible