Liquid AI releases free d1 decision models small enough to make calls in milliseconds on a shop-floor computer
Liquid AI released two open-weight decision models on October 7: d1-3B, which reads text and images, and d1-omni-600M, an early experimental version that reads text with images or audio. Rather than writing text, they return structured answers to set questions in a single pass. Liquid says d1-3B answers a question in 8 ms on an NVIDIA RTX 4090 graphics card and 50 ms on a small Jetson Orin Nano edge board, and scores ahead of every model under 10 billion parameters on its Decision Index benchmark. Both are free to download on Hugging Face.
What it means for manufacturers and fabricators
A model this small can run on a box next to a camera or machine with no internet connection or per-call fees, which makes always-on checks like spotting a missing part or a bad weld on every frame realistic for a mid-sized shop.