Liquid AI is a foundation model company based in Boston, Massachusetts, spun out of MIT.

We build high-performance, efficient AI systems purpose-built for real-world environments —delivering advanced capabilities directly to smartphones, laptops, vehicles, and embedded systems, where compute is limited and efficiency is critical.

Our mission is to build capable and efficient general-purpose AI systems at every scale— models that are not just powerful, but optimized for deployment in compute-constrained, edge environments. Liquid models are engineered for speed, memory efficiency, and real-world reliability, bringing cutting-edge AI closer to where data is generated when privacy, latency, and security are critical.

We take a first-principles approach to model architecture, drawing from deep expertise in dynamical systems, signal processing, and numerical linear algebra. This foundation allows us to design AI systems with unmatched structural control, interpretability, and performance.

Liquid specializes in building best-in-class small language models and general-purpose foundation models that process complex, sequential, multimodal data—text, audio, video, time series, and signals—while maintaining high speed, compact memory footprint, and production scalability. Unlike traditional transformer-based systems, Liquid Foundation Models (LFMs) are architected from the ground up for efficient operation and on-device intelligence.

The name “Liquid” reflects our commitment to creating dynamic, adaptive learning systems—AI that flows with the problem, adapts to its environment, and fits seamlessly within hardware constraints. We work hand-in-hand with enterprise partners to embed Liquid models into mission-critical systems, unlocking new levels of performance, reliability, and efficiency at the edge.