The Kimi K3 model reportedly outperforms Anthropic's Opus 4.8 and is broadly comparable to leading Western models like GPT-5.6 and Fable 5 on benchmarks. This performance challenges the perception that Chinese open-source models consistently lag US closed-source models by six to eight months. Zhipu AI also plans to release a GLM model in the Mythos class by the end of the year.
Chinese AI labs are closing the capability gap with Western frontier models despite sanctions, increasing competitive pressure on US model developers.