Liquid AI has released d1, its first decision model that supports both text and image inputs. Unlike standard language models, d1 processes unstructured data in a single forward pass to return probabilities for structured decisions without generating tokens.

  • Matches or beats GPT-6.1 Sol on four of six real-world applications while costing 19x to 200x less.
  • Answers significantly faster than competitors, with text decisions taking 200 to 300 ms.
  • Supports Noul (yes/no), Choice, and Score question types in a single request.
  • Demonstrates high accuracy in visual inspection tasks without specific training for those domains.

The model is available via the Liquid AI API, Vercel, and OpenRouter, billed on input tokens only.