Liquid AI open-sources d1 decision models for edge to datacenter
- Source
- x.com
- Date
Today we release Open d1: two open-weight multimodal models in our d1 decision model family. > d1-3B: text + vision > d1-omni-600M: text + image or text + audio > Real-time decision making anywhere, from data centers such as @nvidia DGX to RTX workstations to Jetson at the edge. 1/

Context
Liquid AI has released two models from its d1 family, which it calls decision models, according to its announcement on X. Open-weight means the trained files are published, so anyone can download and run them on their own hardware. The larger model, d1-3B, accepts text and images. The smaller one, d1-omni-600M, accepts text together with either an image or an audio clip. Liquid pitches both for real-time decisions on hardware that ranges from NVIDIA DGX systems in data centers, through RTX workstations, down to Jetson devices at the edge.
The stored post covers only what the models take as input and where Liquid says they can run. It reports no results, and it does not explain how a decision model differs from a chat model that writes out its answers. Liquid's earlier d1 announcement, covered in a separate Intel item, describes what the family is built to return. The post is marked as the first part of a thread, but no later parts were stored.
- Open weights — A model whose learned parameter values have been made available to download, under terms the publisher sets.
- Model — A computational system that uses learned patterns to produce an output from information supplied to it.
- Benchmark — A fixed set of tasks with a stated way of scoring them, used to compare models or setups.
- multimodal — A model that works with more than text — reading images, audio, or video, and sometimes generating them too.
Checking sign-in…
Loading comments…






