Perceive Beating AI Newsflash: The SmartVision released its flagship light model GLM-5.3 Flash to benchmark against DeepSeek V4 Flash, utilizing over 100,000 domestic chip clusters for this model's inference service.

Reportedly, its computing power may come from Huawei, Moore Threads, or SeaLight. The SmartVision has not commented officially on this information. In its technical documentation, SmartVision stated that its cluster's "hardware efficiency and unit token cost have reached a level equivalent to mainstream NVIDIA GPUs," but did not specify the exact model. (Stay tuned)