[Kimi K2.8 Preview Fully Launched: Performance Approaching K3, Supports 1M Context] Beating AI Newsflash: The dark side of the moon has begun fully rolling out Kimi K2.8 Preview to Kimi Code CLI. Users do not need to switch Model IDs, as the original kimi-for-coding will directly point to the new model. Kimi states that the overall performance of K2.8 Preview is already close to K3, but with higher thinking efficiency. It also supports the same thinking intensity adjustment as K3 and expands the context to 1M tokens. The previous generation K2.7 Code only supported 256K context and could only enable thinking. This time, K2.8 essentially brings down two core capabilities from K3 to the default coding model. [Original Link]