Liquid AI

Liquid AI represents a breakthrough in foundation model efficiency, creating AI models that deliver maximum intelligence with minimum compute requirements. As an MIT spin-off founded by leading researchers, Liquid AI has pioneered novel neural network architectures called Liquid Foundation Models (LFMs) that are purpose-built for speed, efficiency, and real-world deployment across any hardware environment. Unlike traditional foundation models that require massive computational resources, LFMs are optimized to run seamlessly on GPUs, CPUs, and NPUs, making high-capability AI accessible on devices ranging from wearables and smartphones to laptops, cars, and enterprise servers. The platform offers comprehensive solutions from custom AI development for enterprises to developer tools for building specialized models. Liquid AIs unique architecture enables models to maintain excellent performance while using significantly less memory and compute than comparable models, making them ideal for edge deployment and cost-sensitive applications. The company provides enterprise solutions through device-aware model architecture search, allowing rapid development of custom models optimized for specific hardware constraints and business requirements. For developers, Liquid AI offers LEAP, a platform for building, specializing, and deploying on-device AI, along with Apollo, a mobile app for testing small language models directly on phones. The models support multiple modalities including text, audio, vision, and multimodal capabilities, with parameter sizes ranging from 350M to 1.6B parameters optimized for different use cases and deployment targets.