On the evening of November 6, Moonshot AI unveiled its open - sourced thinking model, Kimi K2 Thinking. This model showcases significantly enhanced agent and reasoning abilities.

One of its remarkable features is that it can think while utilizing tools. This enables the model to autonomously conduct up to 300 rounds of tool calls and engage in multi - round thinking processes without any user input, thus effectively tackling complex problems.

In benchmark tests like 'The Last Human Exam' and BrowseComp, Kimi K2 Thinking delivered outstanding performances, reaching SOTA (State - of - the - Art) levels. Moreover, the model has witnessed notable advancements in areas such as creative writing, academic research, and personal emotional responses.

Users have the opportunity to experience this model either through kimi.com or the Kimi App. Additionally, its API service has been launched simultaneously, providing more accessibility for developers and users alike.