on-device inference
A real model,
running in this tab.
Instruction-tuned weights downloaded once and executed on your GPU through WebGPU. Nothing is sent anywhere — no key, no server, no telemetry.
checking webgpu
- parameters
- 1.5B
- weights
- 1.79 GB · q4
- context
- 32K tokens
- vocabulary
- 151K
- backend
- wasm
- licence
- apache 2.0