Liquid AI is an MIT-spun-out foundation model company building efficient AI systems for real-world deployment across smartphones, laptops, vehicles, and embedded systems. Its core focus is delivering capable general-purpose intelligence in environments where compute, memory, power, latency, privacy, and security constraints matter.
Liquid AI takes a first-principles approach to model architecture and specializes in compact, high-performance foundation models that span text, vision, and audio workloads. Alongside its model portfolio, Liquid AI offers developer tooling and solution engagements that help customers customize, evaluate, and deploy AI locally, in the cloud, or in hybrid environments.
Offerings, Capabilities, and Integrations
Liquid AI combines model development, customization, and deployment tooling for organizations that need AI to run efficiently outside traditional data-center settings. Its capabilities center on low-latency inference, compact memory footprints, multimodal processing, and hardware-aware deployment across CPUs, GPUs, and NPUs.
Liquid AI supports both self-service and guided adoption paths. Developers can test, customize, and deploy models through its tooling, while enterprises can engage Liquid AI for data generation and curation, model training and evaluation, hardware optimization, and production deployment. Its ecosystem also extends into cloud and hardware partnerships, enabling access to models through AWS and cloud partners such as Together AI and Modal, as well as optimized execution with silicon and runtime partners including AMD, Intel, Qualcomm, ExecuTorch, Ollama, LM Studio, and Nexa AI.
Products and Services
- Liquid Foundation Models: Liquid AI’s flagship family of efficient foundation models for text, vision, and audio workloads, designed for deployment on device, in the cloud, or in hybrid environments.
- Liquid Edge AI Platform (LEAP): A developer-first platform for customizing, evaluating, and deploying Liquid Foundation Models across devices and operating systems, with support for on-device workflows and rapid productionization.
- Liquid Apollo: A mobile app for trying Liquid Foundation Models directly on a phone and experiencing private, low-latency on-device AI interactions.
- Liquid Playground: A web-based environment for testing Liquid AI models and experimenting with model behavior before deeper customization or deployment.
- Enterprise Solutions: Custom AI engagements for enterprises that need tailored models and deployments, including data generation, model training, evaluation, hardware optimization, and production support.
- Startup Solutions: A startup program that gives selected venture-backed companies access to Liquid AI’s model stack, LEAP, documentation, community resources, and direct guidance from product and engineering teams.
- LFM2.5: A newer Liquid Foundation Models family for on-device AI spanning text, vision-language, and audio-language models, built for reliable agents and efficient edge deployment.
- LFM2-24B-A2B: Liquid AI’s largest LFM2 model, designed to run from cloud environments to AI PCs and mobile devices while maintaining strong efficiency characteristics.
- LFM2-VL-3B: An efficient vision-language model aimed at multimodal edge use cases that require low-latency visual understanding on constrained hardware.
- Liquid Nanos: A family of highly compact, task-specialized foundation models built for agentic workloads such as extraction, tool use, multilingual translation, and retrieval-augmented generation on everyday devices.
- LFM2-2.6B-MMAI: A specialized scientific foundation model developed for pharmaceutical research workflows and positioned for deployment in research environments.
Target Customers
Liquid AI targets organizations and developers that need AI to operate where latency, privacy, reliability, and hardware efficiency are business-critical. That includes enterprises deploying AI in products and workflows, software developers building on-device applications, and startups looking to create differentiated AI-native offerings without depending entirely on large cloud models.
Its customer focus is especially strong in industries where local or hybrid inference creates clear value, including automotive, financial services, e-commerce, consumer electronics, healthcare and life sciences, and industrial or robotics environments. Liquid AI also serves teams building for mobile devices, laptops, vehicles, embedded systems, and other edge endpoints where traditional large-model deployment is impractical.
Cloud Integrations and Marketplace
- AWS Marketplace: Liquid AI maintains an AWS Marketplace presence with seller listings and deployable SageMaker model packages, including Liquid LFM 40B offerings.
- Together AI: Liquid AI provides access to LFM2-24B-A2B through Together AI for serverless, production-oriented cloud deployment.
- Modal: Liquid AI makes LFM2-24B-A2B available through Modal for rapid customization and self-deployment in cloud environments.
Key People
- Ramin Hasani: Co-founder & CEO
- Mathias Lechner: Co-founder & CTO
- Alexander Amini: Co-founder & CSO
- Daniela L. Rus: Co-founder
- Jeffrey Li: Chief Operating Officer
- Paul Sieminski: Chief Legal Officer
- Rachel Link Robinson: Chief People Officer
- Jessica Sabert: Head of Sales
- Erika Wool: Global Head of Partnerships
- Maxime Labonne: Head of Post-Training
Key Facts
- Headquarters: Cambridge, Massachusetts, United States
- Employees: Approximately 121
- Annual Revenue: Undisclosed
- Parent Company: None
- Subsidiaries: None
- Publicly Listed: No (privately held)
