Moonshot released Kimi K3, a 2.8 trillion-parameter open-weight model claimed to perform near Anthropic's Fable, following Z.ai's GLM-5.2, which reportedly scored on par with top closed-source US models. The reports signal material compression of the performance gap between Chinese and US frontier models, a shift attributed to accelerated release cycles and cost-driven competition. This challenges the standing consensus that Chinese models lag US frontier by six months or more and may compress the addressable-market advantage of US vendors' closed-source moat, particularly if open-weight alternatives reach parity on benchmark measures.
326 other tracked developments touching the same claims.