Kimi K2 Thinking set new records in benchmark assessments evaluating reasoning, coding, and agent capabilities.

Jin10 Data, November 8 — According to the Moon’s Dark Side official website, Kimi K2 Thinking has set new records in benchmark evaluations for reasoning, coding, and agent capabilities. K2 Thinking achieved a state-of-the-art score of 44.9% in the HLE benchmark, 60.2% in the BrowseComp test, and 71.3% in the SWE-Bench Verified test, demonstrating the strong generalization ability of the most advanced thinking agent model.