Liquid AI launches Liquid Labs research unit
Liquid AI has formally named its research team Liquid Labs, committing it to fundamental work on efficient, adaptive AI and to publishing open technical reports, ablations and model weights.
Liquid AI announced Liquid Labs, a formal name for the research group the company says has been central to its work from the start. The team's stated goal is "fundamental breakthroughs in the science of building intelligent, personalized, and adaptive machines."
The company traces its origins to MIT CSAIL, where work on Liquid Neural Networks produced a class of dynamical, efficient sequence-processing architectures. That research became the basis for Liquid Foundation Models (LFMs), which Liquid AI describes as efficient multimodal models for edge, enterprise and scientific use. According to the company, LFMs draw on dynamical systems, efficient sequence-processing mechanisms and hardware-aware co-design, with an emphasis on efficient scaling laws, hardware-adaptive designs and "transparent and analyzable behavior."
Liquid Labs will work across architecture and model design, training objectives, data generation, algorithmic advances and inference optimization. The company says it will continue to publish technical reports, architectural deep dives, ablations, evaluations and model weights. It framed this as a commitment to AI progress that is "transparent, reproducible, and shared."
The announcement comes days after Liquid AI released its technical report on LFM2. The report covers the hybrid architecture, training strategies and benchmarking behind the company's claim that LFM2 models are state-of-the-art in efficiency and speed.
The post gives no headcount, leadership, budget or specific research roadmap for the new unit. It closes with a recruiting call.