Know What's Happening in Technology

Liquid AI Releases LFM2.5: A Compact AI Model Family For Real On Device Agents

Liquid AI unveiled LFM2.5, a compact 1.2 billion‑parameter model family designed for on‑device and edge inference. The suite includes Base, Instruct, Japanese, vision‑language, and audio variants, all released with open weights on Hugging Face and via the LEAP platform. Pre‑training was expanded to 28 trillion tokens, and the Instruct model incorporates supervised fine‑tuning, preference alignment, and multi‑stage reinforcement learning, delivering benchmark scores that outpace competing 1 B‑class models. Specialized variants achieve state‑of‑the‑art results on Japanese, multimodal, and speech tasks while maintaining low latency on CPUs and NPUs.