China's moonshot claims its Kimi K2 thinking beats GPT-5
Chinese AI lab Moonshot just dropped Kimi K2 Thinking, and they're saying it beats top models like OpenAI's GPT-5 and Anthropic's Claude Sonnet 4.5.
It's aced tough benchmarks—think Humanity's Last Exam, BrowseComp, and Seal-0—showing off strong reasoning skills and web info gathering.
While its coding is solid, what really stands out is how well it tackles complex problems using different tools.
Kimi K2 is free and open-source
Kimi K2 is not only advanced, but also open-source and totally free on Hugging Face—built for just $4.6 million (way less than US rivals).
This could shake up the AI world by giving businesses a powerful tool without the usual price tag.
There are some security concerns since it's from China, so companies will need to balance those risks with all the new possibilities this model brings.