LiquidAI: d1

LiquidAI d1 is a structured decision model from Liquid AI. It is listed on api14 under the model id liquid/d1 and the provider LiquidAI, and its main category is decisions. Rather than generating free form prose, d1 is built around making choices, which makes it a different kind of tool from a general chat or writing model. If your application needs a model to pick an option, assign a score, or answer yes or no, that is the role d1 is designed to fill.

The model takes text as its input and returns decisions as its output. Its context window is 65536 tokens, and the maximum output is 58982 tokens. The provider describes d1 as a structured decision model served as a System One endpoint: you send a state along with typed questions, and it returns a choice, a score, or a yes/no answer. That framing matters when you plan an integration, because the response you get back is meant to be consumed by code rather than read as conversational text.

The practical use cases follow directly from that shape. Anywhere a system has to evaluate a state and commit to an answer, d1 can serve as the decision step: routing a request down one of several paths, scoring an item against criteria you supply, or returning a boolean for a gate or filter. Because the input is text and the output is a decision, it fits naturally into pipelines where an upstream component assembles context and a downstream component acts on the result. It is not positioned as a generator of long explanations, so treat it as the component that decides, not the one that narrates.

You can call it through api14's OpenAI-compatible API using the model id liquid/d1. Because the endpoint follows the OpenAI chat completions format, existing clients and SDKs work without special handling. One thing to check before you build: d1 lists no supported request parameters, so do not assume that sampling controls or similar options will have an effect. Send your state and typed questions in the request, and read the decision out of the response.

The main constraint to keep in mind is the 65536 token context window. If your task depends on very long inputs, plan how you will fit them, since the state and the questions both draw on that budget. And if what you actually need is text generation, summarisation, or open ended conversation, a model in a different category will fit better than d1. Choose d1 when the thing you want back is a decision.