01 / The short version
What happened
The d1-3B and d1-omni-600M decision models, built on Liquid Foundation Models, deliver fast single-pass responses for structured decisions. d1-3B achieves a mean score of 82.9 across seven public datasets, outperforming Decider 4B, while d1-omni-600M scores 78.4 with a quarter of the parameters of Decider 2B. d1-3B retains vision capabilities from its LFM2.5-VL-3B backbone and runs in under 50 ms on edge devices and under 10 ms on GPU. Both models are open-weight and available on Hugging Face, with d1-3B requiring trust_remote_code=True for loading. No vision or audio benchmarks are reported due to private or open problem status in Decision Index v0.3.
See the exact references02 / Key takeaways
What you need to know
- 01
d1-3B scores 82.9 mean on seven public datasets, exceeding Decider 4B.
- 02
d1-omni-600M scores 78.4 with only a quarter of Decider 2B's parameters.
- 03
d1-3B answers in under 50 ms on edge devices and under 10 ms on GPU.
Keep in perspective
What to watch for
The source does not report vision or audio benchmarks, noting they are either private splits or open problems in Decision Index v0.3.
Go to the source
Exact references
These are the original pages used for this brief. Publisher claims are not independent evaluations.
01Primary source · Hugging FaceRead the original announcementhttps://huggingface.co/blog/LiquidAI/open-d1AI-generated from the linked source. It can miss context; verify consequential details in the original. How the radar works