Home › Pricing
Kimi K3.1 Pro Pricing
Kimi K3.1 Pro costs $2 per million input tokens and $10 per million output tokens on the official Kimi API — about 33% cheaper than Kimi K3 on both sides. Below are the official rates, what they work out to for real workloads, and how third-party providers compare.
Figures on this page are taken from the official Kimi release notes published 23 October 2026, cross-checked against the Kimi API platform the following day. Last synced: 24 October 2026. Rates change without notice — verify on the official pricing page before you commit to a budget.
Official Kimi API price per 1 million tokens
Kimi K3.1 Pro bills purely per token: you pay for what you send (input) and what the model generates (output), with no monthly platform fee and no minimum spend. The 1M-token context window does not carry a long-context surcharge.
| Model | Input / 1M | Output / 1M | Context | vs K3 |
|---|---|---|---|---|
| Kimi K3.1 Pro | $2.00 | $10.00 | 1M | −33% |
| Kimi K3 | $3.00 | $15.00 | 1M | baseline |
| Kimi K3.1 Pro Highspeed | $4.00 | $20.00 | 1M | +33% |
Highspeed is the same K3.1 Pro model served on a priority tier at exactly double the standard rate on both input and output. Against Kimi K3 ($3 / $15) that premium works out to +33%, the mirror image of the standard tier's −33%.
One structural point worth knowing: output is five times the price of input. That ratio, not the headline token price, is what decides most real bills — a chat product that streams long answers pays far more than a summarisation service that reads long documents and returns a paragraph.
What it costs for real workloads
List prices per million tokens are hard to reason about, so here are three worked examples. Each one compares the same workload on Kimi K3.1 Pro and on Kimi K3.
| Workload | Tokens in / out | K3.1 Pro | K3 | Saved |
|---|---|---|---|---|
| Batch summarisation of 1,000 documents | 10M / 2M | $40.00 | $60.00 | $20.00 |
| Chat assistant, 1,000 sessions | 1M / 0.5M | $7.00 | $10.50 | $3.50 |
| Codebase-wide refactor pass | 100M / 10M | $300.00 | $450.00 | $150.00 |
Because both sides of the price dropped by exactly one third, the saving is a flat 33% at any input-to-output ratio — no workload shape is penalised. If you are migrating from K3, expect your bill to fall by a third at identical traffic.
Kimi K3.1 Pro token cost calculator
Enter your monthly token volume in millions to see what it costs on Kimi K3.1 Pro, what the same volume cost on Kimi K3, and the difference.
Calculator uses the list rates $2 / $10 (K3.1 Pro) and $3 / $15 (K3) per million tokens. It does not model cached input, batch discounts or provider markups.
Kimi K3.1 Pro pricing at third-party providers
Beyond the official API, K3.1 Pro is served by third-party inference platforms. Their rates vary because each one sets its own margin, routing policy and caching rules, and some bundle free credits on signup. The official list price is the anchor to compare against:
| Provider | Input / 1M | Output / 1M | Note |
|---|---|---|---|
| Kimi official API | $2.00 | $10.00 | Reference rate, no middleman |
| OpenRouter (standard) | $2.00 | $10.00 | Same as official; cache reads $0.20 |
| OpenRouter (highspeed tier) | $4.00 | $20.00 | Priority service tier, 2× standard |
| Requesty gateway (pay as you go) | $2.10 | $10.50 | Official rate + 5% routing markup |
| Requesty gateway (BYO keys) | $2.00 | $10.00 | No markup when you supply the key |
Router and gateway rows are derived from their published markup over the upstream Moonshot rate (Requesty: 5% pay-as-you-go, 0% with your own key), not from a separate quoted price list. Verify on the provider's page before committing budget.
One caveat that catches people out: a cheaper per-token rate is not automatically a cheaper month. Providers differ in throughput, retry behaviour and max context, and a provider that truncates a 1M-token request into several smaller calls can charge you more than the official rate for the same task. Benchmark on your own workload before moving production traffic.
Kimi subscription versus API pricing
These are two different products and they are not interchangeable. The API is metered per token and is what you budget for with the table above. The consumer subscription is a flat monthly fee for using Kimi through the app and web interface, with usage caps set by the plan. A subscription does not give you API keys at zero marginal cost, and an API account does not include consumer-app access.
Subscription tiers and their limits are announced on the official site and change often, so this page does not restate them. If you are choosing between the two, the deciding question is simple: are you building something that calls a model (API), or are you a person doing work in a chat window (subscription)?
How to reduce your Kimi K3.1 Pro bill
1. Move work from output to input
Output tokens cost five times what input tokens cost. Constrain answer length with explicit instructions and max-token limits, and restructure tasks so the model edits or selects rather than regenerates. Cutting output volume by half typically saves more than any provider shopping.
2. Reuse context instead of resending it
The 1M-token window makes it tempting to paste the whole codebase into every call. That is the fastest way to burn budget. Keep a stable prefix, cache what the platform lets you cache, and send only the delta. (Cached reads bill at $0.20 per 1M — see the caching note above.)
3. Match the model to the task
Not every request needs a 3T-parameter model. Route classification, extraction and short rewrites to a smaller model and reserve K3.1 Pro for long-context reasoning. Mixed routing routinely cuts cost by more than a price cut does.
4. Meter per feature, not per account
Log token spend by endpoint from day one. Without per-feature metering you will discover which feature is expensive only after the invoice, which is exactly when it is most expensive to fix.
Frequently asked questions
How much does Kimi K3.1 Pro cost per million tokens?
On the official Kimi API platform, Kimi K3.1 Pro is priced at $2 per million input tokens and $10 per million output tokens. There is no minimum commitment and no separate charge for long-context requests.
Is Kimi K3.1 Pro cheaper than Kimi K3?
Yes. Kimi K3 was priced at $3 per million input tokens and $15 per million output tokens. At $2 and $10, Kimi K3.1 Pro is about 33% cheaper on both the input and the output side, so the saving holds whatever your input-to-output ratio is.
How much does it cost to process 10 million tokens with Kimi K3.1 Pro?
It depends on the input-to-output split. Sending 10 million input tokens and receiving 2 million output tokens costs $40 on Kimi K3.1 Pro ($20 for input plus $20 for output), against $60 on Kimi K3. Ten million tokens split evenly between input and output would cost $60 ($10 plus $50).
What is the difference between Kimi K3.1 Pro and Kimi K3.1 Pro Highspeed?
Highspeed is a lower-latency serving variant of the same K3.1 Pro model, offered for workloads where time to first token matters more than list price. It is priced separately from the standard rate, so compare the two rates against your actual latency requirements before migrating a production workload.
Do third-party providers charge the same as the official Kimi API?
No. Third-party inference providers and API resellers set their own per-million-token rates, and they can sit above or below the official $2 / $10 list price depending on routing, caching and margin. Always compare the effective per-million price, not the headline discount.
Is there a free way to use Kimi K3.1 Pro?
The official consumer app typically offers a limited free allowance that is separate from API billing, and the model weights are published, so self-hosting is possible if you have the hardware. Neither route gives unlimited production API capacity at zero cost, and free tiers change frequently, so check the official platform before relying on one.
Next: how the model actually performs — see the Kimi K3.1 Pro benchmarks, or the overview of specs and release details.