Purpose
Liquid AI's Decision Models are specialized models built for structured decision-making scenarios, focusing on three types of tasks: classification, routing, and scoring. Their characteristic is that they directly produce a decision result in a single call, and do not generate any tokens.
Main Capabilities
According to the material, Decision Models provide three output forms:
- Noul: A yes/no question that returns a probability between 0 and 1. For example, "Is this message spam?" might return 0.92.
- Choice: A question that selects one option from multiple choices, returning a probability distribution over several named options.
- Score: A question that scores according to a scale, returning a probability-weighted position on an ordered rubric.
How to Use
The material page provides a documentation index, and the complete documentation index can be obtained at /llms.txt, for discovering all available pages before browsing further. The page also includes two topic entry points: "When to Use Decision Models" and "Build a Game with Decision Models".
The material specifically notes that field names, SDK coverage, preprocessing, and confidence semantics differ between models. The documentation tables require horizontal scrolling to view on small screens.
Applicable Scenarios
Suitable for tasks that require making structured judgments, such as classification, request routing, and scoring according to established criteria. In the material, "Is this message spam?" is used as an example of Noul, and it is mentioned that Decision Models can be used to build a game.
Limitations and Information Not Provided
The material does not explain the models' specific parameter scale, deployment method, invocation interface details, the specific languages or platforms supported by the SDK, pricing and billing methods, performance metrics, or accuracy data. The above content is not currently provided. In addition, the material does not give specific integration steps other than the documentation index entry point.
