Kimi K2: A New Chinese AI Outperforms ChatGPT and Cloud

Achieves Impressive Results in Key Benchmarks

Kimi K2: A New Chinese AI Outperforms ChatGPT and Cloud

Chinese company Moonshot AI has achieved a remarkable feat with its new Kimi K2 Thinking model, which outperformed both the American ChatGPT and Anthropic's Cloud in several important benchmark tests.

Since the beginning of this year, the landscape has clearly shifted following the significant success of the Chinese company DeepSec, which demonstrated that it is possible to build powerful AI models comparable to Western ones like ChatGPT and Gemini, but at a much lower cost, according to a report published by Livemint. Since then, major Chinese companies such as Alibaba (Coin), Baidu (Erny), and Moonshot AI (Kimi) have entered the race aggressively, making the competition for the top spot closer than ever.

Kimi K2 Beats ChatGPT and Claude

According to data published by Moonshot, the Kimi K2 model scored 44.9% on the Humanity’s Last Exam (HLE), outperforming GPT-5 (High) at 41.7% and Claude Sonnet 4.5 Thinking at 32%. In another test, BrowseComp, which evaluates models' performance in web browsing and self-searching, Kimi again outperformed GPT-5 with a score of 60.2%, compared to 54.9% for GPT-5 and 24.1% for Claude.

However, in competitive programming tasks, ChatGPT's dominance remains evident. Kimi scored 83.1% on the LiveCodeBench V6 test, compared to 87% for GPT-5 and 64% for Claude. In LMArena’s rankings, Gemini 2.5 Pro remains in the lead, while GPT-5 comes in fourth, ahead of Kimi.

Free and available to everyone

The Kimi K2 Thinking model is now available on Kimi’s official website and app on the Apple and Android app stores. It is one of the few models that users can try for free without any message limits. In contrast, OpenAI imposes strict restrictions on users of the free version of GPT Chat, denying them access to the GPT-5 (High) model and limiting the number of messages to just 10 every 5 hours.

With this achievement, China demonstrates that it is no longer merely trying to catch up with the West in artificial intelligence, but has become capable of surpassing the strongest global models in some aspects. With increasing competition, the dominance of American companies in this sector appears to be facing a real challenge from the East.