Home / Compare / Kimi K2.6 vs Kimi K3

Kimi K2.6 vs Kimi K3

For a typical chat reply (2,000 tokens in, 400 out), Kimi K3 costs 4.5× as much as Kimi K2.6: $12.00 against $2.66 per 1,000 requests at list price on 2026-09-28.

Kimi K2.6Kimi K3
Input, per 1M tokens$0.65$3.00
Output, per 1M tokens$3.41$15.00
Context window262K1.05M
Classification or routing (600 tokens in, 10 out), per 1,000$0.424$1.95
Chat reply (2,000 in, 400 out), per 1,000$2.66$12.00
Retrieval-augmented answer (8,000 in, 500 out), per 1,000$6.91$31.50
Confidence signalyesyes
Tool callingyesyes
BANKING77, correct of 400323not measured

Which one should carry your traffic?

If Kimi K2.6 answers your requests as well as Kimi K3, moving that traffic saves 78% on chat-shaped requests. Averages cannot tell you whether it does for your requests. Kimi K2.6 exposes a confidence signal, so Gimbix can let it answer first and hand unsure cases to Kimi K3.

Kimi K2.6 profile · Kimi K3 profile · Measure it on your traffic