On-device AI deployment has reached a major breakthrough. The open-source project Edge0 has released two models—Edge0-145B and Edge0-120B—along with a unified inference framework and deployment tools. Edge0-145B is claimed to be the world's first open-source model with 145 billion parameters that can run smoothly on a personal computer locally. The team innovated across four dimensions—underlying architecture, training, loading, and scheduling—adopting technologies such as sparse expert architecture, on-device collaborative training, a three-tier storage intelligent pipeline, and efficient CPU-GPU collaboration. These innovations dramatically lower the barrier to operation, achieving low memory usage, high generation speed, and ultra-long context support on mainstream consumer PCs. Its core capabilities are comparable to cutting-edge cloud-based large models and can support various complex local workflows offline. Edge0 has built a complete model matrix covering multiple modalities, and both model weights and deployment tools have been open-sourced, driving a shift in the AI application paradigm toward an on-device-first, cloud-assisted approach.