Models - Liquid AI Docs

Pass a model’s id as model in your request. Your organization’s models, with their rate limits and prices, are listed on the Models page of Liquid Console, and GET /openai/v1/models returns the ones you can call. See the Liquid AI pricing page for current prices.

Generation models

Call these through chat completions.

Model	Input	Output	Context window	Best for
`lfm3-1.7b`	Text	Text	128,000	Fast agents, tool use, and everyday chat.
`lfm2.5-24b-a2b`	Text	Text	128,000	Our largest model: a reasoning mixture of experts (24B, 2B active) for harder tasks.
`lfm3-vl-2b`	Text, images	Text	128,000	Vision: describing, reading, and reasoning about images.
`lfm3-audio-2b`	Text, audio	Text, speech	128,000	Audio: transcription, translation, and spoken replies.
`lfm3-omni-2b`	Text, audio, images	Text, speech	128,000	Audio and images in one conversation.

Decision models

Call these through the decisions endpoint. They answer typed questions with calibrated probabilities and never generate tokens.

Model	Input	Context window	Best for
`d1-ultra`	Text, images	32,768	The most capable decision model, for harder decisions.
`d1.1`	Text, images	65,536	The default decision model.

Open-weight models

Most of these models are also open-weight, so you can run them yourself on laptops, phones, and GPUs. See Liquid Foundation Models for every model and how to run it.