Liquid AI releases LFM3, small open models for on-device agents

Liquid AI released LFM3, a family of six open-weight models in the 2B-parameter class covering text, vision, audio, an omni model that sees, listens and talks, and decision models, all built on one shared base. The text model fits in about 1 GB at 4-bit precision, has a 256K-token context window and is tuned for tool calling and local agent harnesses. Liquid says LFM3-1.7B beats its older 2.6B model with 35% fewer parameters, and that every model can be fine-tuned on a single GPU. The results are Liquid's own; the release includes Colab notebooks for jobs like pulling JSON from messy documents and reading receipt photos.