LFM3 is the third generation of Liquid Foundation Models, published by Liquid AI on October 27, 2026. Liquid describes it as a starting point for specialized intelligence: one shared base that developers can adapt into the models they need.
The release brings text, vision, audio, omni and decision models onto a single shared backbone, all in the 2B parameter range. It ships six 2B-class models trained from LFM3-1.7B-Base, including Liquid's first omni model, and the vision model now accepts video input.
According to Liquid, LFM3-1.7B outperforms LFM2.5-2.6B while using 35% fewer parameters. The text model fits in about 1 GB at 4-bit precision and offers a 256K context window, so agents can run directly on the device.
Liquid's architecture search covered 14 device and backend combinations, from phones and a Raspberry Pi 5 to laptops, cloud CPUs, GPUs and the Qualcomm Hexagon NPU. Every model can be fine-tuned on a single GPU, which keeps the bar low for teams that want a small model inside their own product.