Settings:

Model size:

Small Medium Big

Small: 200MB; Medium: 400MB; Big: 1.4GB. Big will probably crash this website.

Models are cached in your browser and run locally (when available with webgpu).

Infos:

Using SmolLM2 Models and Transformer.js.

External Models:

gemma2-9b-it llama-3.1-8b-instant deepseek-r1-distill-llama-70b

The models are hosted on Groq and run on their hardware. Their Privacy Policy applies.