Kimi K2.8 Preview Fully Launched: Performance Approaches K3, Supports 1M Context
1 hour ago / Read about 0 minute
Author:小编   

Moonshot AI has fully rolled out the Kimi K2.8 Preview version to Kimi Code CLI. Users no longer need to switch Model IDs, as the original kimi-for-coding will now directly point to the new model. The overall performance of K2.8 Preview is close to that of K3, with improved thinking efficiency and adjustable thinking intensity. The context window has been expanded to 1M tokens. This version integrates two core capabilities of K3 into the default code model, whereas the previous generation K2.7 Code had a context window of only 256K and only supported the thinking function.