Chinese AI lab Moonshot AI is preparing to release Kimi K3, an open-weight model reportedly expected to match or even exceed the performance of Anthropic's Opus 4.8, according to the Financial Times citing anonymous sources. This matters because it would mark another significant step by a Chinese developer to close the performance gap with leading closed-source Western AI labs, adding fuel to an ongoing industry debate about whether it is worth paying for expensive proprietary models when cheaper open-source alternatives are catching up.

Kimi K3 is expected to be the largest open-weight AI model produced in China, with a parameter count of between two and three trillion, and is reportedly due for release within days. It follows Moonshot's well-received Kimi K2 models, which already rank highly on benchmarks. The company is also said to be in talks to raise fresh funding at a valuation of $31.5 billion (£24.6bn), up from the $20 billion valuation it achieved after raising $2 billion in May. The report comes amid growing concern among industry figures that AI labs could exploit client data submitted to closed products such as ChatGPT and Claude, prompting some executives to promote cheaper open-source alternatives from firms including DeepSeek, Z.ai and Moonshot.