LFM3

Model Oct 28, 2026

Liquid AI's compact open-weight family for local text agents, vision, audio, multimodal interaction and structured decisions.

Visit primary link 3 sources attached

Liquid AI's LFM3 architecture diagram shows convolution, full attention and sliding-window attention blocks across 24 layers. — Author-published architecture diagram from the English LFM3 launch guide. It explains the backbone design, not an independently reproduced performance result. Source: Liquid AI

self-hosted multimodal ai-agents decision-models

What happened lately

LFM3 timeline

Launches Oct 28, 2026

Liquid AI's LFM3 brings eyes, ears and tool calls to a compact local-model family

Developing 4 sources

Liquid AI announced the LFM3 family on October 27, 2026. Its shared compact backbone supports separate text, vision, audio, omni and decision checkpoints. The text model targets local tool use and structured output; the company explicitly discourages treating it as a model for complex agentic coding. The published text checkpoint uses the LFM Open License v1.0, including a commercial revenue threshold, rather than an unrestricted permissive license.