LFM2-1.2B
Liquid AI · 2025-07-10 · 1.2B parameters
LFM2-1.2B is Liquid AI's 1,170,340,608-parameter post-trained hybrid language model, part of the second-generation LFM2 family designed for low-latency on-device and edge deployment. Its 16-block architecture combines 10 double-gated short-range convolution blocks with six grouped-query attention blocks, and its pre-training ran for 10 trillion tokens; it supports a 32,768-token context window.