Kimi Code's K2.7 Code HighSpeed mode is out of beta and now a permanent option. The high-speed variant uses the same model as the standard version, but cranks up output to roughly 180 tokens per second in typical coding scenarios, and up to 260 tokens per second with short contexts.
Allegretto and above subscribers can use it immediately without requesting access — just fire up the Kimi Code CLI, type /model, and select kimi-for-coding-highspeed. It also works in any tool that supports the Coding Plan. The catch: high-speed mode burns through tokens at triple the rate of the standard version, so it's best for tasks where responsiveness matters more than budget.