Tech Desk 
Nov 10, 2025
A screengrab taken on November 10. — YouTube/@Sam Witteveen
Competition in the realm of artificial intelligence (AI) has intensified with the introduction of Kimi K2 Thinking, a new model from Beijing's Moonshot AI Lab claimed to outperform OpenAI's GPT-5 and Anthropic's Claude Sonnet 4.5 on various leading benchmarks.
The most notable part of the Chinese Kimi K2 AI model, which truly gives it an upper hand over GPT-5 and Claude Sonnet 4.5, is that it is available for free as it is open-source.
Kimi K2 Thinking is defined as a reasoning-focused, Mixture-of-Experts (MoE) model that is designed for complex academic and analytical tasks.
The latest Chinese AI model integrates long-horizon planning and adaptive reasoning, alongside online tools like web browsers.
When it comes to tests like Humanity’s Last Exam and BrowseComp, it excelled in reasoning and planning tasks and came on par with its competitors in coding performance without significantly outperforming them.
Kimi K2 is trained on approximately 1 trillion parameters and is openly accessible on Hugging Face, allowing developers to bend the code to their will.
The arrival of Kimi K2 challenges the upward trend of subscription-based AI platforms, with the developer, Moonshot, claiming a development cost of just $4.6 million, substantially lower than that of leading US AI labs.
Experts are of the view that Kimi K2 could disrupt traditional business models, prompting companies to reconsider expensive AI subscriptions in favour of powerful, free alternatives.
That bends the competitive scale in Kimi's favour is that some US firms, like Airbnb, are already exploring Chinese-developed models for better performance and cost efficiency.
