Chinese Unveils Kimi K2-Thinking AI Model

Credit: Pixabay

The Chinese company Moonshot AI has begun deploying a reasoning version of its artificial intelligence model, dubbed Kimi K2-Thinking. The developer has published test results showing that the model performs on par with advanced American models.

How to test. The easiest way is to use the web version of the Kimi chatbot. You need to register with a Google account. After logging in, select the K2 model in the query input field and enable the Thinking option.

Please note that when using the mobile app, Kimi may request confirmation via SMS code, but you won’t be able to enter a Russian phone number.

What it can do. According to Moonshot AI, the model performs better than or on par with GPT-5 and Claude Sonnet 4.5 on a number of benchmarks, including Humanity’s Last Exam, which contains 2,500 questions from various fields.

Kimi K2 Thinking was listed on Hugging Face. It claims to have 1 trillion parameters and 32 billion active parameters.

It is also available through the API at $0.60 and $2.50 for 1 million input and output tokens, respectively, and $1.15 and $8 in Turbo mode.