× LFM3

A conversation that thinks, on your own CPU

Liquid AI's LFM3 models, compiled to WebAssembly with fusor-ai: LFM3-1.7B, a text chat model; LFM3-VL-2B, which reads images too; LFM3-Audio-2B, which hears speech and speaks; and LFM3-Omni-2B, which does all of it; and d1-2B and d1-Omni-2B, decision models that answer named, typed questions about a state and its images (and, for d1-Omni-2B, its clips) in one pass. Ask anything, show it an image, play or record it a clip, let it reason in a thinking block first or answer straight away, and read or hear the reply as it is written. Once loaded, nothing you type, show or say leaves the page.