[Tencent Hunyuan Releases Hyra 2.0, Driving AI Development from Solving Scientific Problems to Independently Developing Scientific Models and Tools] According to DeepFlow TechFlow on November 2nd, Tencent Hunyuan officially released its AI research system, Hyra 2.0, shifting its focus from directly solving scientific problems to independently developing and upgrading reusable scientific models, research workflows, and software tools. The new system adopts a three-tiered Agent architecture (Overall, Mission, Worker), supporting long-term task management, parallel experiments, independent review, and research experience accumulation. Tencent Hunyuan disclosed that Hyra 2.0 has independently trained the protein design model HyInv and the protein language model HyPLM-Flash. HyInv's sequence generation speed is 57 times faster than ProteinMPNN; HyPLM-Flash has 650 million parameters and boasts the fastest inference speed among 12 compared models. In addition, Hyra 2.0 has made progress in areas such as quantum error correction, robot structure design, and chip development, including increasing the geometric mean of the SPEC CPU2006 score per GHz of the open-source RISC-V processor Xiangshan by 5.79%. These results will be gradually made available for community verification and reuse.