AliCloud Tongyi open-sources first multimodal inference model QVQ, visual inference comparable to OpenAI o1

On December 25, AliCloud Tongyi Qianqian released the industry's first open source multimodal reasoning model, QVQ-72B-Preview.QVQ demonstrated super-expected visual comprehension and reasoning capabilities, and excelled in solving complex reasoning problems in the fields of mathematics, physics, and science. A number of evaluation data show that QVQ surpasses the previous visual understanding model Qwen2-VL, and the overall performance is comparable to OpenAI o1, Claude3.5 Sonnet and other reasoning models. Currently, developers can experience it directly on the Magic Ride community and HuggingFace platform. (36Krypton)