China's New AI Giant, Moonshot AI, Challenges U.S. Dominance

According to the company, the model is on par with the top models from its U.S. competitors, OpenAI and Anthropic. The launch is causing a stir in the industry and fueling concerns in the U.S. about the country's dwindling technological lead.

The "largest open-air model" in the world

With 2.8 trillion parameters, Kimi K3 is, according to the company, the world’s largest open-source AI model to date. The number of parameters is an indicator of a system’s performance and complexity. The model is set to be made fully available to developers worldwide by the end of July 2026.

During the presentation, Moonshot AI cited test results showing that Kimi K3 outperforms models such as Anthropic’s Claude Opus 4.8 and OpenAI’s GPT 5.5, at least when it comes to programming tasks and complex applications. These test results could not be validated by an independent third party.

OpenAI and Anthropic's lead is shrinking

At the same time, Moonshot AI acknowledged that, overall, the new system still lags slightly behind the absolute, proprietary top U.S. models—namely Claude Fable 5 and OpenAI’s brand-new GPT-5.6 Sol—but that the gap is closing rapidly. Observers are already drawing parallels to the so-called “DeepSeek moment” in early 2025, when a Chinese model shook assumptions of U.S. dominance for the first time and caused tech stock markets worldwide to plummet.

Chinese AI models have since gained prominence worldwide, in part because many of them are made available as free, open-source software and are therefore generally more cost-effective to operate than Western AI versions.