Liquid AI launches multimodal decision model d1-ultra, scores 65 on Artificial Analysis index

Liquid AI has introduced d1-ultra, which it describes as its most powerful decision model to date. The model can take text, images, or both as input and answers constrained questions in a single forward pass. Instead of generating output tokens, it assigns probabilities to each possible answer, making it suited for classification, routing, and scoring workloads. Liquid AI said developers can use the model for these tasks without paying for text generation. d1-ultra is now available through the Liquid API at $0.042 per million input tokens. The company also said the model is built on a larger open-weight backbone than its predecessor, d1, with extensive custom improvements. In an internal test suite covering 112 evaluations, d1-ultra led across all eight categories. According to a Liquid AI report, the model ranked first on the independent Artificial Analysis Decision Index with a score of 65. The update was cited by MarkTechPost and carried by Techub News.