Moonshot AI

Direct Moonshot AI language and reasoning models.

3 integrated models · Official documentation ↗

3 models

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.

Language ModelsJun 15, 2026262,144 context

$1.615 / 1M tokens Input$6.80 / 1M tokens Output

Kimi K2.7 Code

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.

Language ModelsJun 12, 2026262,144 context

$0.8075 / 1M tokens Input$3.40 / 1M tokens Output

Kimi K2.6

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 Superseded as flagship by Kimi K3.

Language ModelsApr 20, 2026262,144 context

$0.8075 / 1M tokens Input$3.40 / 1M tokens Output

Page 1 of 1