Short answer: Moonshot’s API changelog lists kimi-k2-thinking and kimi-k2-thinking-turbo as retired and no longer supported on that platform; the excerpt does not specify regional scope. A supplied OpenRouter listing excerpt reports a NovitaAI route for non-Turbo kimi-k2-thinking at $0.60 per million non-cached input tokens, $2.50 per million output tokens, and $0.15 per million cache-read tokens. Treat these as listed third-party route rates, not Moonshot-direct prices or a confirmed live quote. OpenRouter’s pricing documentation identifies its model-price fields as USD. Source check for this revision: November 10, 2026; the model listing itself was not independently refreshed for this revision. (openrouter.ai)
The supplied listing excerpt does not establish a current rate for kimi-k2-thinking-turbo. Moonshot’s 2025 announcement gave Turbo rates effective November 6, 2025, but those historical prices do not establish current availability or pricing.
Rates by exact model ID
| Model ID | Moonshot API status | Rate information in the supplied sources | Important qualification |
|---|---|---|---|
kimi-k2-thinking |
Moonshot’s API changelog says the model is retired and unsupported on that platform; the excerpt does not define regional scope. | OpenRouter’s supplied listing excerpt reports NovitaAI rates of $0.60/M non-cached input, $2.50/M output, and $0.15/M cache read. | Third-party route rates, not Moonshot-direct pricing. USD is supported by OpenRouter’s pricing documentation; confirm the live listing and provider terms before deployment. |
kimi-k2-thinking-turbo |
Moonshot’s API changelog says the model is retired and unsupported on that platform; the excerpt does not define regional scope. | No current rate established here. The 2025 announcement listed $0.15/M cache-hit input, $1.15/M cache-miss input, and $8.00/M output. | Historical announcement rates effective November 6, 2025—not confirmation of current pricing or an available route. |
The supplied OpenRouter excerpt says requests for the listed non-Turbo model are forwarded to its listed provider. That does not establish Moonshot-direct support, nor does it confirm that the route supports every feature needed by a particular application.
Illustrative one-month token subtotal
For an illustrative 30-day month using the listed non-Turbo route, assume 7 million non-cached input tokens, 3 million cache-read tokens, and 1 million billable output tokens:
- Non-cached input: 7 × $0.60 = $4.20
- Cache read: 3 × $0.15 = $0.45
- Output: 1 × $2.50 = $2.50
- Conditional token subtotal: $7.15
This is an example based on assumed token volumes, not a quote or complete production budget. It assumes the 3 million cache-read tokens qualify for the listed cache-read rate. The 1 million output-token assumption should include any reasoning tokens that the route bills as output; OpenRouter’s general documentation says reasoning tokens are considered output and charged accordingly, but the supplied evidence does not confirm NovitaAI’s route-specific accounting. (openrouter.ai) Other charges and whether a Kimi membership benefit applies to this third-party route are not established here.
Production availability and tool support
Moonshot’s changelog lists both Thinking variants among the retired K2 series and directs users to migrate to Kimi K3. This is what the cited Moonshot API-platform notice says; its regional applicability is not established in the supplied excerpt.
The supplied OpenRouter listing identifies a NovitaAI route for non-Turbo kimi-k2-thinking, but the available evidence does not confirm tool or function-calling support, route-specific restrictions, or provider limits. Moonshot’s 2025 launch announcement described the Thinking models as intended for complex reasoning, multi-step instructions, and agent-like tasks such as tool use and function calling. That launch description is not proof of current tool support on the NovitaAI route.
Bottom line
- Non-Turbo via the listed OpenRouter/NovitaAI route: supplied listing rates are $0.60/M non-cached input, $2.50/M output, and $0.15/M cache read. Treat them as reported listing prices, and confirm current terms before production use.
- Moonshot-direct API: the cited changelog says both Thinking variants are retired and unsupported on that platform; regional scope is not specified in the excerpt.
- Turbo: $0.15/M cache-hit input, $1.15/M cache-miss input, and $8.00/M output are announcement-era rates effective November 6, 2025. A current Turbo rate or route is not established here.
- Budgeting: separate non-cached input, cache-read input, and output; include any billed reasoning tokens in output. The example above is a conditional token subtotal, not a complete budget.
Sources
Platform Changelog — Kimi API Platform
Pricing — OpenRouter Documentation