Liquid AI releases open-weight d1-3B and d1-omni-600M decision models

Liquid AI has released two open-weight multimodal decision models: d1-3B (3.12B parameters, text and images) and d1-omni-600M (587M, text with image or audio). Instead of generating tokens, they return calibrated probabilities for named questions in one forward pass; the license allows free commercial use below $10M annual revenue.

• d1-3B scores 48.57 on Decision Index v0.2.1, the best result under 10B parameters
• One question takes 8 ms on an RTX 4090 and 16 ms on Jetson AGX Thor
• d1-omni-600M accepts up to 30 seconds of audio or an image, never both
• Both models report output_tokens: 0 and are on Hugging Face with llama.cpp support