AI that runs
on your machine.
Pick a model below. It downloads once, caches locally, and runs entirely in your browser. No servers, no API keys, no limits.
How it works
Three simple steps to run AI locally. No cloud, no setup, no cost.
Pick a model
Choose from text, vision, or training models. They download once and cache locally in your browser.
Run locally
Inference runs entirely on your device via WebGPU or WASM. No data ever leaves your machine.
Own your data
No servers, no API keys, no limits. You control everything — the model, the data, the privacy.
Text
Vision
Training
Create your own personality
Fine-tune SmolLM2-360M on any text using LoRA + Unsloth. Export to ONNX, upload to the Chat page, and start talking to your custom AI — all in your browser.
1. Train
Use your GPU or Google Colab. ~5-15 minutes.
2. Export
Automatic ONNX export with --export-onnx flag.
3. Upload
Drag the ZIP into the Chat page.
4. Chat
All inference runs locally via WebGPU.
100% local
Your data stays on your device. No uploads, no servers, no third parties.
Zero cost
No API fees, no GPU rentals, no subscriptions. Your hardware does the work.
Offline-ready
Works after the first download. No internet required for inference.