Liquid AI

Liquid AI describes itself as an efficiency-first foundation model company. Its stated mission is efficient, general-purpose AI at every scale, so that models run on processors outside data centres, where latency, privacy and hardware limits matter. The company is Liquid AI, Inc., a Delaware corporation whose terms are governed by Massachusetts law.

What it does

The products are Liquid Foundation Models, called LFMs. The home page reports 60 LFMs shipped, more than 5,100 variants and 54.2 million downloads, with models for text, vision and audio. Recent releases listed include LFM3, announced on 27 October 2026, LFM2.5 models for on-device agents, a vision-language model for the edge, and a family of decision models called d1. The models run through llama.cpp, MLX, ONNX, Ollama, SGLang, vLLM and other runtimes.

Customers can use models three ways. A hosted API through the Liquid Console, OpenRouter or the Vercel AI Gateway is billed per million tokens, and each new organisation gets $2 of free credit. Open weights can be downloaded, run, fine-tuned and kept private under the LFM Open License, with no copyleft. And for enterprises, Liquid builds and optimises custom models for a deployment environment, with OEM and on-prem support. The company says it never uses customer data to train models and that tailored models deliver up to five times better memory use or speed than adapting a general open model.

Who it serves

The site lists deployments in phones, laptops, cars, space, e-commerce, financial services, biotech and defence. Named customers and partners include Shopify and Mercedes-Benz, and MacPaw and Qualcomm are announced as partners.

Key facts

Legal name: Liquid AI, Inc. (Delaware corporation) Origin: spun out of MIT CSAIL Founders: Ramin Hasani (CEO), Mathias Lechner (CTO), Alexander Amini (CSO), Daniela Rus Open weight licence: free for companies under $10M annual revenue; research, education and non-profit use always free; paid enterprise licence above $10M Hosted API example: LFM2.5-24B-A2B at $0.05 per million input tokens and $0.70 per million output tokens Support: support@liquid.ai