Moonshot AI has introduced its inaugural Kimi model bearing the 'Pro' designation, K3.1 Pro. Leveraging the robust foundation of K3's 2.8 trillion parameter architecture, this new model substantially bolsters its Agent functionalities through extensive reinforcement learning techniques. In a series of evaluations, K3.1 Pro showcased exceptional prowess, securing the premier position among open-source models, despite its 3D capabilities yielding mixed outcomes. Notably, the API Tokens for K3.1 Pro are now priced at one-third less than those of K3, albeit with a twofold increase in the consumption of input Tokens, and a roughly 15% rise in the cost per query during network defense assessments. Presently, K3.1 Pro is accessible across multiple platforms, with the complete model weights slated for release no later than November 4. To preempt potential misuse, the official team has preemptively excised certain network attack and defense training datasets. Vulnerability exploitation tests indicated that the success rate of the final iteration was inferior to that of the control version, which lacked risk mitigation protocols.