According to CoinMeta, Tencent's Hunyuan has released Research agent Hyra 2.0, which expands the focus of research from repeatedly trial-and-erroring on individual problems to developing scientific models, algorithms, and tools that can be reused over the long term. Hyra 2.0 divides the work among three layers: overall, mission, and worker, which are responsible for resource coordination, research planning, and experiment execution, respectively. Tencent disclosed that Hyra 2.0 has trained two protein models from scratch; among them, the model with 1.96 million parameters, Hyinv, can design protein sequences based on target skeletons, and its generation speed is 57 times faster than that of commonly used proteinmpnn. The other model, Hyplm-flash, has 650 million parameters and utilized 726 billion training token. Hyra 2.0 has also contributed to optimizing the open-source RISC-V processor "Xiangshan," resulting in a 5.79% increase in performance per GHz while only increasing the chip area by 0.3%. Currently, Hyra 2.0 is also involved in the pre-training, post-training, and inference system optimization of the next-generation Hunyuan model, attempting to improve AI through enhanced development tools to help train an even stronger next-generation AI.