High-speed serving of Kimi K2.7 Code: the same coding model streamed at about 180 output tokens per second, up to 260 on short inputs, with thinking always on and a 256K context.
$1.615 / 1M tokens Input$6.80 / 1M tokens Output
3 models
Coding-specialized Kimi built on K2.6 for long-horizon software engineering with fewer thinking tokens; open-weight 1T MoE (32B active), 256K context, thinking always on. Sits below Moonshot's flagship Kimi K3.
$0.8075 / 1M tokens Input$3.40 / 1M tokens Output
Moonshot's open-weight general model from April 2026: a 1T-parameter MoE (32B active) with both thinking and non-thinking modes over a 256K context. Superseded as flagship by Kimi K3.
$0.8075 / 1M tokens Input$3.40 / 1M tokens Output
Page 1 of 1