Liquid AI Releases Open-Weight d1-3B and d1-omni-600M: Multimodal Decision Models With Zero Output Tokens

Liquid AI has introduced Open d1, adding two open-weight models to its decision model family. The d1-3B handles text and image inputs, while the smaller d1-omni-600M processes text paired with either images or audio.

Unlike conventional language models that generate text token by token, these decision models produce calibrated, typed answers in a single forward pass. They output zero text tokens, positioning them for classification and structured decision tasks rather than conversational generation.

The open-weight release gives developers access to compact multimodal models that can reason across text, images, and audio without the latency and cost of autoregressive decoding. This could benefit real-time applications where fast, structured outputs matter more than free-form text.

By offering a 600M-parameter model that handles audio alongside text and vision, Liquid AI targets edge and resource-constrained deployments. The approach challenges the assumption that useful AI must always generate language, expanding the toolkit for task-specific multimodal inference.

This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.