The Kimi Code CLI is free; access to the models behind it is sold in three ways — a Kimi membership starting at $19 per month, an Extra Usage wallet that takes over when your membership quota runs out, and the Kimi Open Platform pay-as-you-go API, priced at $3 per million input tokens and $15 per million output tokens for K3.The exact membership quota figures are not published, so choosing among the three depends on usage; the calculation below shows where the break-even point falls. Prices read from kimi.com and platform.kimi.ai on October 1, 2026.
To learn what Kimi Code is and how it compares with other tools, see Kimi Code vs Claude Code; to switch clients, memberships, or models, see Kimi Code alternatives. The English original of this page is Kimi Code pricing.
Membership plans
| Plan | Monthly billing | Annual billing | Kimi Code and K3 | K3 1M context, HighSpeed |
|---|---|---|---|---|
| Plus | $19 | $15 per month ($180 per year) | Yes | No |
| Pro | $39 | $31 per month ($372 per year) | Yes | Yes |
| Max | $99 | $79 per month ($948 per year) | Yes | Yes |
| Ultra | $199 | $159 per month ($1,908 per year) | Yes | Yes |
Prices are from the kimi.com membership page (the prices are rendered in the browser and do not appear in the page source); the two columns on the right are from Kimi's membership documentation. Free accounts and the Go plan do not include Kimi Code quota. Kimi recently renamed the plans with “no price change”; existing plans retain their old names and rules until you switch. All prices are in US dollars; the actual amount charged in New Taiwan dollars for users in Taiwan depends on the exchange rate and the card issuer at the time of payment.
How the quota works
- Everything shares one pool. Requests from the CLI, VS Code, desktop app, and third-party tools all count against the same membership quota, shared across every signed-in device and API key.
- A rolling five-hour window limits bursts and recovers automatically. New members no longer have a weekly limit; legacy plans still reset every seven days from the subscription date, with no rollover.
- A total monthly quota is shared with Kimi on the web. Once it runs out, Kimi Code is frozen until reset or upgrade.
- Figures for all three are unpublished — neither token counts nor request counts. You can check your remaining quota in the Kimi Code Console or in the CLI with
/usage.
Extra Usage: the wallet beyond your quota
When your membership quota runs out, Extra Usage continues paying for requests from a prepaid balance. The rules are also from the membership documentation:
- Only members with an active subscription can enable it, and Kimi on the web and Kimi Code share the same balance.
- Billed in yuan according to actual usage, at rates Kimi describes as close to Open Platform API prices — Kimi itself also says that upgrading the subscription is usually better value for heavy use.
- It is charged last, after the subscription quota, so tasks keep running instead of failing. Without a monthly spending limit set in settings, there is no limit.
- Each top-up must be at least ¥25; the daily maximum is 10 top-ups and ¥3,000, and the balance cap is ¥10,000. The balance does not expire and remains after a subscription lapses (but you cannot top up while it is lapsed); it is generally non-refundable.
Pay-as-you-go API pricing per token
| Model ID | Input (cache miss) | Input (cache hit) | Output | Context |
|---|---|---|---|---|
kimi-k3 — Cached input: $3.00 (5 minutes) or $6.00 (1 hour) | $3.00 | $0.30 | $15.00 | 1,048,576 tokens |
kimi-k2.7-code | $0.95 | $0.19 | $4.00 | 262,144 tokens |
kimi-k2.7-code-highspeed | $1.90 | $0.38 | $8.00 | 262,144 tokens |
Per million tokens, before tax, from platform.kimi.ai; the API base URL is https://api.moonshot.ai/v1 and it accepts requests in OpenAI and Anthropic formats. This is a different product from the membership: a separate account and key, and model IDs (kimi-k3, kimi-k2.7-code) that differ from the membership IDs (kimi-for-coding, k3). Membership cannot pay for API calls, and API balance does not increase membership quota.
Membership or pay-as-you-go: where is the break-even point?
The following is token arithmetic, not measured task cost. One session sends 200,000 uncached input tokens and receives 12,000 output tokens:
| Route | One session | Equivalent to this many Plus ($19) sessions | Who bills you |
|---|---|---|---|
kimi-k3, pay-as-you-go | $0.78 | About 24 times | Kimi Open Platform |
kimi-k3, 80% of input cache hit | $0.35 | About 55 times | Kimi Open Platform |
kimi-k2.7-code, pay-as-you-go | $0.24 | About 80 times | Kimi Open Platform |
| Claude Haiku 4.5 through Kunavo | $0.18 | About 104 times | Kunavo |
| Claude Sonnet 5 through Kunavo | $0.36 | About 52 times | Kunavo |
Read this as a threshold. If you run fewer K3 sessions per month than the number in the third column, the pay-as-you-go API costs less than Plus; if you run more, Plus may be better value — until the five-hour window or total monthly quota stops you, which is exactly what Kimi has not quantified. The cached row excludes the first cache-write charge ($3 or $6 per million, depending on cache lifetime). “Lowest unit price” and “cheapest way to finish the task” are different things: a model that needs three retries may cost more than a pricier model that gets it done in one.
The two Kunavo rows are the current price list unit prices for Claude models; Kimi Code can connect through the openai provider type, but Kunavo does not sell Kimi models. The price-list amounts are billing floors, not ceilings — when the upstream provider reports a cost, the bill uses the higher of the price-list amount and the upstream cost multiplied by the applicable markup. Kunavo uses prepaid top-ups and charges per token, with a minimum top-up of $10 and no subscription (billing details); checkout is through Stripe, with Taiwanese cards (including JCB and UnionPay), Apple Pay, Google Pay, and Link available in Taiwan; JKoPay and LINE Pay are not on the available list. The provider configuration block is on Kunavo's Kimi Code settings page. No one at Kunavo has actually run Kimi Code against its own endpoint, so after the first request, check usage records to see the actual amount deducted.
Frequently asked questions
How much does Kimi Code cost?
Kimi Code CLI itself is free MIT-licensed software (MoonshotAI/kimi-code, version 2.1.1 released September 24, 2026). The model usage is what costs money, and there are three ways to pay. First is Kimi membership, which includes Kimi Code starting with Plus: as of October 1, 2026, kimi.com lists Plus at $19 per month ($15 per month with annual billing, $180 per year), Pro at $39 ($31, $372), Max at $99 ($79, $948), and Ultra at $199 ($159, $1,908). The second is Extra Usage, which takes over once membership credits are used up and is topped up in Chinese yuan. The third is to skip membership and pay per token on the Kimi Open Platform: kimi-k3 costs $3 per million input tokens and $15 per million output tokens.
Which plan do I need for Kimi Code and K3?
Plus is the starting point: Kimi's membership documentation says Kimi Code and K3 are available from Plus upward, while K3's 1M-token context and HighSpeed require Pro or above. The Go plan has no coding quota. Higher plans increase the Agent quota on kimi.com (the pricing page lists Pro at 2×, Max at 5×, and Ultra at 10× relative to Plus), but no plan publishes specific token or request limits for Kimi Code.
What happens when I run out of Kimi Code quota?
New members have two limits. The first is a rolling 5-hour rate window, which limits short-term bursts and resets automatically; weekly limits have been removed for new members, while legacy plans still reset every 7 days. The second is a total monthly quota shared with the Kimi web app; once it runs out, Kimi Code is frozen until the quota resets or you upgrade. After that, Extra Usage lets you continue making requests with a prepaid balance in Chinese yuan, at rates Kimi says are close to Open Platform API prices. If you enable Extra Usage without setting a monthly spending limit, there is no cap.
Is Kimi membership cheaper than the Kimi API?
For heavy use, probably — Kimi's own documentation says that extensive Extra Usage is less cost-effective than upgrading your subscription. The deciding factor is usage. At the Kimi Open Platform price for kimi-k3, a session with 200,000 uncached input tokens and 12,000 output tokens costs about $0.78; Plus at $19 is roughly equivalent to 24 such sessions per month. With kimi-k2.7-code, it is about 80 sessions. Kimi has not published how many sessions Plus actually allows, so the honest way to judge is: if your monthly pay-as-you-go bill would exceed the plan price, try the plan and check usage in Kimi Code Console.
Can Kimi Code use other or cheaper APIs?
Yes. Kimi Code's providers documentation describes a provider type called openai, which lets you change base_url to any OpenAI-compatible service; the configuration is in ~/.kimi-code/config.toml. The /login menu itself only offers Kimi Code OAuth or a Kimi Platform API key, so other providers require editing the configuration file. The membership and Open Platform have different model IDs and endpoints, so a configuration written for one cannot be applied directly to the other. Kunavo does not sell Kimi models; its Kimi Code configuration page points the openai type to Claude and GPT IDs.
Read on October 1, 2026: kimi.com/membership/pricing (rendered in the browser), Kimi Code membership documentation, the chat pricing and quickstart pages on platform.kimi.ai, and the GitHub API for MoonshotAI/kimi-code. No Kimi account was used, and nothing was purchased or run; all amounts are calculations based on published rates. Since membership quotas are unpublished, no per-token plan price is calculated.