Liquid AI

Liquid AI is an MIT CSAIL spinout building foundation models based on novel neural network architectures called Liquid Neural Networks, which go beyond standard transformers by combining structured state-space, convolutional, and attention mechanisms. Their flagship LFM (Liquid Foundation Models) series — including the hybrid MoE LFM2-24B — delivers competitive performance with lower compute and memory requirements. The company targets enterprise and edge deployment use cases where inference efficiency is critical.