Kimi K2 Thinking
The November 2025 reasoning version of Kimi K2, built to keep thinking across hundreds of chained tool calls without losing the thread. Good for research and multi-step agent work at a low price, open content level. Text only and slower than plain K2. Kimi K3 is the newer pick.
Kimi K2 Thinking is a legacy model. Its successor, Kimi K3, is faster or cheaper - and chats that used Kimi K2 Thinking move to it automatically.
See Kimi K3The facts
A typical message: 16K tokens in, 1.5K out. Short replies cost less. Models served via OpenRouter bill the real provider cost plus our markup, so actual charges vary. 1 credit = 0.01 €.
Compare Kimi K2 Thinking
Same facts, side by side.
Try Kimi K2 Thinking now.
No subscription. Free credits to start. The chat opens with Kimi K2 Thinking selected and a prompt ready - just send.