Liquid AI has released d1, a decision model designed for structured choices rather than text generation. The model takes context and a set of typed questions, then returns calibrated probabilities across a fixed set of outcomes in a single call. According to the announcement, it does so with zero generated tokens.
This differs from typical large language models, which produce free-form text. d1's output is a probability distribution over predefined options, making it relevant for decision tasks where calibrated confidence is important. The company's description emphasizes the absence of generated tokens, suggesting a fundamentally different inference path.
The source does not provide details on the underlying architecture or training data. It also does not specify the exact target use cases beyond structured decisions. What is clear is that d1 is positioned as a decision model, not a text generator.