Kimi K3 API pricing at a glance
For Kimi K3 (kimi-k3), Kimi’s pricing page lists these rates per 1 million tokens:
| Token type | Listed price |
|---|---|
| Cache-hit input | $0.30 / 1M tokens |
| Cache-miss input | $3.00 / 1M tokens |
| Output | $15.00 / 1M tokens |
These are listed K3 rates—not universal prices for every Kimi model, membership, or third-party provider. Kimi says applicable taxes are excluded and calculated at checkout based on billing region. The pricing page labels itself updated October 27, 2026; that is the page’s displayed date, not confirmation that these rates are currently available to a particular account. (Kimi K3 pricing page)
Production-route qualification: The captured pricing page does not identify which API route or region its USD schedule applies to. Kimi’s separate platform documentation lists an overseas Kimi Platform base URL, but the available material does not connect that route to these K3 rates or confirm their applicability to a US billing account. Treat the figures below as a conditional estimate based on the listed schedule, not a confirmed quote for an overseas production route. (Kimi Code and Platform documentation)
Kimi also lists a separate K3.1 Pro pricing page. Do not assume K3’s rates apply to K3.1 Pro; use the schedule for the exact model and product you intend to call. (Kimi K3.1 Pro pricing page)
Estimate a Kimi K3 API bill
If the listed USD schedule applies to your intended route and account, estimate model charges as:
(uncached input tokens ÷ 1,000,000 × $3.00)
+ (cache-hit input tokens ÷ 1,000,000 × $0.30)
+ (output tokens ÷ 1,000,000 × $15.00)
Illustrative example: 1 million uncached input tokens, 1 million cache-hit input tokens, and 1 million output tokens would total:
$3.00 + $0.30 + $15.00 = $18.30
This is arithmetic from the listed rates, not a measured bill or a quote for a specific route. It excludes applicable taxes and other costs. Kimi describes context caching as automatic and lists a lower rate for cache-hit input, but the pricing page does not specify cache eligibility or retention rules. Don’t assume repeated-looking prompts qualify for the cache-hit rate; use the actual usage categories reported for your API calls. (Kimi K3 pricing page)
API billing versus membership
Kimi’s documentation distinguishes Kimi Code, described as a monthly or annual membership with rate limits, from Kimi Platform, described as pay-as-you-go with top-ups. It directs developers integrating model calls into their own products to Kimi Platform. A membership price should not be treated as the cost of production API traffic; confirm the current billing terms for the product and account you will use. (Kimi Code and Platform documentation)
Kimi K3 capabilities—and what the pricing page doesn’t establish
The K3 pricing page lists a 1,048,576-token context window, automatic context caching, tool calls, JSON mode, structured output, and reasoning control through the top-level reasoning_effort field. These capability statements do not confirm that a particular account can use K3 on a particular production route. (Kimi K3 pricing page)
Before committing a production budget, check current official pricing and account or checkout information for the exact model and route. The available material does not establish whether the listed USD schedule applies to an overseas route or US billing account. Separately confirm model availability, applicable taxes, endpoint and authentication instructions, rate and concurrency limits, and cache behavior in documentation for the product you plan to use. Do not transfer Kimi Code setup details to Kimi Platform unless the documentation explicitly applies to that product.
Sources
Kimi K3 Pricing | Plans, Membership & API Costs