Liquid AI's d1 is available on AI Gateway
Vercel’s AI Gateway now supports Liquid AI’s d1 decision model for structured evaluations, including vision-based classification, via multiple API formats.
Liquid AI’s d1 is a decision model designed to evaluate shared state against typed questions for tasks such as classification, routing, and scoring. Unlike text-generating models, d1 returns structured answers with probabilities, enabling precise decision-making without token generation. The model also supports vision-based queries, allowing users to ask typed questions about images for tasks like visual classification, inspection, and scoring. These capabilities position d1 as a tool for applications requiring deterministic outputs rather than open-ended text responses.
To integrate d1, users can access it through the AI SDK decision API, the OpenAI-compatible Decisions API, or the TypeSafe-compatible API. This flexibility allows developers to incorporate the model into existing workflows using familiar interfaces. The decision model is designed to handle structured inputs and produce consistent, interpretable results, making it suitable for enterprise and production environments where reliability is critical.
Vercel’s AI Gateway now includes d1, providing a unified platform to access the model alongside other AI services. The gateway simplifies model deployment by offering a single endpoint for decision-making tasks, reducing the complexity of managing multiple APIs. Users can track usage and costs for decision calls in the same interface used for other model requests, ensuring transparency and control over expenditures.
Decision calls made through AI Gateway are logged and counted toward usage budgets, similar to other model requests. The documentation outlines support for Choice, Score, and Boolean question types, enabling users to implement a range of decision logic. Additional decision models are available on AI Gateway, expanding the options for structured AI-driven workflows.