By No Code MBA · Facts checked
- Provider
- Liquid AI
- Announcement date
- Status at last check
- Generally available
- Model family
- d1
- Capabilities
- Text, Image, Audio
What changed
d1-3B and d1-omni-600M are Liquid AI's first open-weight decision models, which return yes/no, choice or score answers in a single forward pass instead of generating tokens. d1-3B is built on LFM2.5-VL-3B and takes text and images; Liquid reports a Decision Index v0.2.1 score of 48.57 and 8 ms single-question latency on an RTX 4090. d1-omni-600M uses a bidirectional encoder with vision and audio encoders, taking text plus image or text plus audio.
Access and availability
Download from Hugging Face at LiquidAI/d1-3B and LiquidAI/d1-omni-600M under Liquid's LFM 1.0 license, with day-one llama.cpp support. d1-3B requires transformers 5.14 or later with trust_remote_code.
Official access documentationWhat you can build
Test the d1 models on a bounded decision task using representative text, image, or audio inputs, then compare answer quality and latency with your current model before adopting them.
Limitations
These are not chat models and do not write text. d1-omni-600M is an experimental early checkpoint with no published latency figures, and Liquid says audio decision benchmarks are still an open problem. d1-3B has a 32,768-token context.