Liquid AI
MIT CSAIL spinout building Liquid Foundation Models (LFMs), small hybrid convolution-attention models designed to run on phones, laptops, and other devices without a cloud.
Total Models
1
Text Models
1
Active Period
Aug 2026 to Aug 2026
Founded in 2023 by MIT CSAIL researchers Ramin Hasani, Mathias Lechner, Alexander Amini, and Daniela Rus; based in Boston.
Came out of stealth in December 2023 with a $37.5M seed round, then raised a $250M Series A led by AMD Ventures in December 2024.
LFMs use a hybrid of gated short convolutions and grouped-query attention built for CPU and on-device inference; LFM2 was open-sourced in July 2025.
LFM2.5-2.6B (August 2026): 2.69B parameters, 128K context, always-on reasoning, aimed at agent workflows, tool use, extraction, and RAG rather than coding or knowledge-heavy tasks.