Liquid AI launches LFM3, a family of six small multimodal models
The new tri-hybrid architecture covers text, audio, vision and omni models.
Liquid AI introduced LFM3 on Tuesday, its new generation of Liquid Foundation Models with a tri-hybrid architecture and new multimodal encoders.
The family has six models across text, audio, vision and omni, including a 1.7B text model with a 256k context window, a 2B vision-language model with video input, a 2B audio model with tool calls and voice chat in 18 languages, and a 2B omni model. It also includes decision models with typed outputs and calibrated probabilities.
The company pitches the suite for specialized intelligence. The post drew about 1,400 likes.
