Liquid AI released d1 decision models that answer questions in a single forward pass without generating tokens. d1-3B achieves the best performance under 10B parameters on the Decision Index and supports text and images with inference speeds under 50ms on edge devices. d1-omni-600M is a smaller multimodal variant supporting text, images, and audio.
Fruit Inspector is a real-time conveyor belt inspection system using YOLO object detection and LightGBM classification to grade fruit quality at 40 FPS on standard CPUs. It tracks fruit movement through a control zone on the belt, running inference via ONNX Runtime with support for CPU, OpenVINO, and CUDA backends.