Liquid AI announced d1, its first model built for fast, structured decision-making in software environments. The company says d1 is the first model to outperform Jev on Hugging Face's Decision Index.
Liquid says d1 performs well on multilingual evaluations, is more robust to prompt injection, and handles longer inputs more effectively. Those are the company's launch claims; the announcement does not provide enough methodology or per-task figures to treat them as independently verified results.
The model is available through the Liquid API, and Liquid says it will reach OpenRouter soon. It enters the same emerging category as OpenAI's new Decisions API: models or endpoints designed to choose among structured outcomes inside a larger workflow. The products may differ in interface, cost, and measured performance, which builders should test on their own decision sets.
Source: Liquid AI's announcement
