LiquidAI/d1-3B — new model trending #30 on Hugging Face
Liquid AI releases d1-3B, a 3.1B open decision model that answers structured questions in one forward pass.
Liquid AI released d1-3B, a 3.12B-parameter multimodal decision model built on LFM2.5-VL-3B and trending on Hugging Face. It takes text, JSON, or images plus typed questions and returns calibrated yes/no, choice, or score answers in one forward pass with zero output tokens. Liquid AI reports a Decision Index 0.2.1 score of 48.57, ahead of Decider 35B-A3B at 47.11, and 74.1 on 11 public image benchmarks. It uses a SigLIP2 NaFlex 400M vision encoder, a 32,768-token context, and is quoted at about 8 ms per decision on an NVIDIA RTX 4090.
- 3.12B-parameter model post-trained from LFM2.5-VL-3B with a SigLIP2 encoder.
- Scores 48.57 on Decision Index 0.2.1, claimed best under 10B.
- Returns calibrated typed answers in one pass with zero output tokens.
- About 8 ms per decision on an RTX 4090; 32,768-token context.
- Vendors
- Liquid AI
- Products
- d1-3B
- Organizations
- Liquid AI
- AI models
- d1-3BLFM2.5-VL-3BDecider 35B-A3B
From the original article
7 min read at huggingface.coimage-text-to-text · 5,370 downloads · 135 likes · license other
By LiquidAI · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.