Liquid AI

Liquid AI is an AI Infrastructure & Training foundation model company that builds ultra-efficient multimodal AI models for on-device, cloud, and hybrid deployment, with listed model offers starting at $0 USD and broader production or enterprise pricing handled through custom commercial terms that are not fully published in the provided data.

Founded on 2023-12-06 and spun out of MIT, Liquid AI focuses on Liquid Foundation Models designed for real-world deployment rather than only large cloud inference. The website describes the company as building ultra-efficient multimodal AI models for privacy-critical, low-latency applications, including on-device, cloud, and hybrid environments. Its published model library lists 20 Liquid Foundation Models across text, vision-language, audio, and nano model categories, including LFM2-350M, LFM2-700M, LFM2-8B-A1B, LFM2-24B-A2B, and LFM2.5-1.2B-Base. Several listed model offers show a price of $0 USD and availability as InStock, although the website content provided does not expose a full commercial pricing table for enterprise deployments.

The main value proposition is hardware-aware AI deployment. Liquid AI explicitly targets CPUs, GPUs, and NPUs, which matters for organizations building AI into phones, laptops, cars, embedded systems, private enterprise environments, or other settings where cloud-only models may add latency, cost, or data governance concerns. The model lineup includes compact parameter counts such as 350M, 700M, and 1.2B, as well as larger mixture-of-experts style entries such as 8B-A1B and 24B-A2B, giving teams a range of options when balancing capability, memory footprint, and device constraints.

Compared to the other AI Infrastructure & Training tools in our directory, Liquid AI is most differentiated by its emphasis on efficient multimodal models deployable across edge, cloud, and hybrid settings rather than a general hosted chatbot or API-only assistant. Based on our analysis of 870+ AI tools, it is best evaluated as model infrastructure for teams with deployment constraints, not as a plug-and-play productivity app. Teams that need the broadest model ecosystem, mature developer tooling, or general-purpose hosted assistants may still compare it against OpenAI, Anthropic, Google Gemini, or Together AI; teams that need private, low-latency inference on CPUs, GPUs, and NPUs should put Liquid AI higher on the shortlist.