Chinese lab Moonshot to launch $31.5bn-valued AI model rivalling Anthropic

Chinese AI developer Moonshot AI is preparing to release Kimi K3, a model expected to perform on par with or surpass Anthropic’s Opus 4.8. The launch is slated for the coming days and will introduce the largest open-weight AI model yet developed in China, containing between 2 trillion and 3 trillion parameters.

For European enterprises, the release intensifies a critical strategic choice between expensive US-controlled systems and open-source alternatives. Companies across the continent are increasingly hesitant to feed proprietary data into closed platforms like ChatGPT or Claude due to fears that the American labs behind them could extract and exploit that information.

This anxiety is already reshaping corporate procurement strategies. Executives are actively steering their organizations away from paying premium prices for closed-source models, either by developing in-house alternatives or adopting open-weight architectures from Chinese developers like Moonshot, DeepSeek, and Z.ai. These open models allow companies to train the technology on their own infrastructure without surrendering sensitive corporate data to third parties.

Moonshot’s previous release, the Kimi K2, already established the lab as a serious contender by ranking highly on industry benchmarks despite being open source. The anticipated leap in capability with the K3 model closes the performance gap that previously made closed-source systems the default choice for complex enterprise tasks.

The technological momentum is translating into significant financial backing for the Beijing-based company. Moonshot is currently raising a new funding round that values the business at $31.5 billion. This represents a sharp increase from a $2 billion raise completed in May, which placed its valuation at $20 billion.