The model supports 44 languages and leads all eight categories of Liquid AI's 112 evaluation suite with a calibration error of 1.5. Demos include operational use cases such as identifying broken product listings, picking Pokémon moves, and sorting trash from photos. This release follows the Jev series launch by less than one day, targeting a market for discrete decision-making rather than standard generative chat completions.
Tuesday, October 20, 2026
Story
Liquid AI's d1-ultra Tops Decision Index with No Output Costs
AI5 days ago
The company's new multimodal AI can process images and text to select an answer without generating expensive output tokens. Liquid AI released d1-ultra on Oct. 15, which ranks first on the Artificial Analysis Decision Index with a score of 65, surpassing Jev 1.13 and Jev 1.14, which scored 58 and 57 respectively. Because it only pays for input at a rate of $0.042 per million tokens, the model is designed to lower costs for agents performing binary or multiple choice tasks.
Looking for more context or analysis?
Tilt chat provides reliable, tailored insights, alerts, and automations.