Multimodal open d1 decision models for the edge

Liquid AI has released two new open decision models, d1-3B and d1-omni-600M (experimental), as part of their d1 decision model family. These models are designed for edge applications and support text, vision, and audio. The d1-omni-600M model shows strong performance across various benchmarks, including SQuAD 2.0 (83.3), Civil Comments (95.8), MASSIVE intent (88.3), PubMedQA (68.3), BoolQ (89.0), XNLI (88.6), and PAWS-X (76.4 and 79.5), achieving a mean score of 82.9.

This release marks the first time Liquid AI has offered multimodal open decision models for edge devices, unlike previous models focused solely on text.