The open-weight launch expands on earlier performance data showing that d1 matches or beats GPT-6.1 Sol and Claude Opus 5.5 on 4 of 6 real-world applications. The model utilizes a single forward pass to provide decisions in 200 to 300 ms without generating tokens, which reduces operating costs by 19x to 200x compared to the competitors. For industrial applications, the model achieves 85% to 97% accuracy on VisA inspection tasks.
Tuesday, October 20, 2026
Story
Liquid AI Releases 2 Open Weight Decision Models with Nvidia
AI13 days ago
Liquid AI has launched Open d1, a family of multimodal decision models that can be run on NVIDIA DGX data centers, RTX workstations, and Jetson edge devices. The release consists of two open-weight models: d1-3B, which supports text and vision, and an experimental d1-omni-600M, a 600M parameter model capable of processing text combined with audio or images. The models are designed for real-time, on-device applications, including voice-command routing, on-device moderation, and gesture-controlled games.
Looking for more context or analysis?
Tilt chat provides reliable, tailored insights, alerts, and automations.