Pioneering on-device foundation models that put fast, private AI directly onto the world's devices; from the phone in your hand to the car and the factory floor, no data center required.

Liquid AI is a foundation-model company building a new class of efficient AI designed to run directly on devices. Originating from MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL), the company set out to answer a different question from the rest of the industry: not how large a model can become, but how small it can be while remaining very capable.

At its core is a novel, non-transformer architecture, the Liquid Foundation Models (LFMs), engineered for speed, low memory, and low power consumption. This lets capable AI run locally on phones, cars, PCs, and industrial hardware, delivering fast, private, offline intelligence without depending on the cloud. The same design philosophy extends from language to vision and audio, allowing a single efficient model family to serve many on-device use cases.

Liquid's models have been downloaded more than 42 million times, and partners including AMD, Samsung, Mercedes-Benz, Shopify, G42, and Insilico Medicine are building on them. As more than two billion processors ship every year running almost no real intelligence, Liquid is positioned to become the intelligence layer for the physical world, running on the device, and belonging to whoever put it there.

‍