Moonshot AI fully launched its Kimi K2.8 Preview model on Kimi Code on September 11, keeping the Model ID unchanged as kimi-for-coding so clients and third-party tools can continue using it without configuration changes. The company said K2.8 Preview delivers overall performance close to K3 with higher thinking efficiency. Coding and agent capabilities have been broadly improved from K2.7 Code, and all membership tiers can access a maximum 1M-token context window. The model supports low, high, and max thinking effort levels, with max as the default.