Introducing LFM3: Every Modality for Specialized Intelligence

Liquid AI has released LFM3, a new family of small open-weight models available now on Hugging Face. Six models of around 2 billion parameters share the same LFM3-1.7B base: text, vision (images and video), audio, a first “omni” model that takes text, images, video and audio and can reply in speech, plus decision models. Liquid AI says LFM3-1.7B beats LFM2.5-2.6B with 35% fewer parameters, fits in about 1 GB at 4-bit and handles a 256K-token context. Each model can be fine-tuned on a single GPU to specialise it and then run on-device.

See original publication↗Liquid AI (blog officiel) · liquid.ai