Chat with LFM.
Liquid AI's LFM3-1.7B — a hybrid convolution + attention model built for the edge — answers right here, streamed token by token from your GPU via Hugging Face and Transformers.js. The first message downloads ~1.6 GB once; after that it is cached. Nothing you say leaves the device.
Type below to load the model and start the conversation.
The model keeps the last few turns of the conversation. Your chat history stays in this browser (localStorage) and is never uploaded.