Moonshot AI Kimi K2.8 Preview Model Fully Launched, Performance Approaches K3

TechFlow reports that on September 11, Moonshot's Kimi K2.8 Preview model was fully rolled out on Kimi Code. The Model ID remains kimi-for-coding, allowing client and third-party tools to operate without any configuration changes. Official sources state that the model's overall performance closely approaches that of K3, with upgraded coding and agent capabilities, and significantly improved reasoning efficiency compared to K2.7 Code. Kimi K2.8 Preview supports the same thinking effort adjustment levels as K3, offering low, high, and max settings (defaulting to max), and all membership tiers are granted access to a context window of up to 1M.