Chinese startup Moonshot AI has unveiled its new open-source model, Kimi K2 Thinking, and claims that it outperforms GPT‑5 and Claude Sonnet 4.5 in key benchmark tests. According to the company, Kimi K2 Thinking—built using a mixture-of-experts architecture with a trillion parameters—achieved top scores on reasoning, coding and tool-use evaluations.
Released on 6 November 2025, the model scored 60.2% on the BrowseComp benchmark, ahead of GPT-5’s 54.9%. On coding tests like LiveCodeBench v6 it achieved 83.1%, again exceeding its rivals.Moonshot says Kimi K2 Thinking was developed at a fraction of the cost of some frontier Western models and is now publicly accessible with its weights released under a permissive license.
While independent verification of all claims is still pending, analysts say this could signal a major shift in the global AI arms race, particularly because of the model’s open nature and cost efficiency.
See What’s Next in Tech With the Fast Forward Newsletter
Tweets From @varindiamag
Nothing to see here - yet
When they Tweet, their Tweets will show up here.
