Short answer: As of November 10, 2026, Kimi’s API model list marks kimi-k2.5 as offline, and the current official pricing table does not list a K2.5 rate. That means there is no verified current production API price to quote for Kimi K2.5. Don’t substitute pricing for Kimi K2.6, K2.7 Code, or the older Kimi K2 model: those are different model entries. (platform.kimi.com)
Kimi’s pricing documentation says API calls are billed by input and output tokens, with separate pricing by model and feature. It defines 1 MTok as 1,000,000 tokens. The pricing table currently lists rates for other models, but not K2.5. (kimi.com)
A current-price example for an alternative—not Kimi K2.5
If you need a budget figure for a model that is listed on the current rate card, Kimi K3.1 Pro is priced at ¥13 per MTok of uncached input, ¥68 per MTok of output, and ¥1.30 per MTok of cache-hit input. Cache writes are listed separately: ¥13 per MTok for a 5-minute cache and ¥26 per MTok for a 1-hour cache. These are K3.1 Pro rates, not K2.5 rates. (platform.kimi.com)
Illustrative calculation: Assume one API call uses 1,000,000 uncached input tokens and generates 100,000 output tokens, with no cache writes or add-on features:
- Input: 1.0 MTok × ¥13 = ¥13
- Output: 0.1 MTok × ¥68 = ¥6.80
- Estimated token charge: ¥19.80
If that same 1 MTok of input qualifies as cache-hit input instead, its input charge would be 1.0 × ¥1.30 = ¥1.30, making the token total ¥8.10, before any cache-write or feature charges. This is an example for K3.1 Pro only; actual cost depends on metered usage and applicable billing categories. (platform.kimi.com)
Other billing details to keep separate
Kimi’s billing guide says a call that uses web search has an additional ¥0.03 charge per call, separate from token charges. It also says cache-hit tokens receive a discounted rate; consult the model’s pricing entry for the applicable cache categories rather than assuming all cached or repeated input is charged at the hit rate. (kimi.com)
The current model list also says Kimi K2.5 is offline. Separately, Kimi’s pricing page says K2-series models stopped being available to newly registered users on October 31, 2026, with a planned shutdown of all K2-series models at the end of December 2026. That notice does not provide a K2.5 price or establish that offline K2.5 access remains available to existing users. (platform.kimi.com)
Is Kimi membership a substitute for API pricing?
No. Kimi describes Open Platform API use as token-billed, and its official pricing page distinguishes API inference rates from consumer memberships. A consumer subscription price is not a Kimi K2.5 API rate and cannot be used to calculate an application’s token bill. (platform.kimi.com)
Bottom line: For Kimi K2.5, the verified answer is no current rate found on Kimi’s official pricing table, and the model is listed as offline. For a production budget, confirm the exact model available to your API account and use that model’s row in the official rate card. The K3.1 Pro calculation above is only a transparent alternative-model example—not a K2.5 estimate.