Overview
Moonshot AI makes the Kimi models, and its Open Platform at platform.kimi.ai is the international API for them. The price list covers Kimi K3 at $3 in and $15 out per million tokens, Kimi K2.6 at $0.95 / $4, and K2.7 Code in a standard and a high-speed variant. Cache-hit input is billed separately at roughly a tenth of the cache-miss price.
The account is prepaid. You need to recharge at least $1 before the first call, and reaching $5 of cumulative recharges earns a $5 voucher. Rate limits are tied to cumulative recharge: the $1 tier allows a single concurrent request, 3 requests per minute and 1.5M tokens a day, while the $10 tier lifts that to 15 concurrent requests, 100 RPM and no daily token cap.
Plan for the tier structure when testing: the entry tier is too tight for load tests or agents running in parallel, so a $10 recharge is the practical starting point.
Pricing and plans
| Plan | Price | What you get |
|---|---|---|
| kimi-k3 | $3 in / $15 out per 1M | Cache-hit input $0.30 |
| kimi-k2.6 | $0.95 in / $4 out per 1M | Cache-hit input $0.16 |
| kimi-k2.7-code | $0.95 in / $4 out per 1M | Cache-hit input $0.19 |
| kimi-k2.7-code-highspeed | $1.90 in / $8 out per 1M | Cache-hit input $0.38 |
| Tier 0 ($1 recharged) | $1 minimum | 1 concurrent request, 3 RPM, 500K TPM, 1.5M tokens/day |
| Tier 1 ($10 recharged) | Prepaid | 15 concurrent, 100 RPM, 2M TPM, no daily cap; higher tiers at $20, $100, $1,000, $3,000 |
What it does
- Recharge-based tiers
- Six tiers from $1 to $3,000 cumulative recharge raise concurrency, RPM and TPM.
- Context caching
- Cache-hit input on kimi-k3 costs $0.30 per 1M versus $3 for a cache miss.
- Coding models
- kimi-k2.7-code and a high-speed variant at double the price.
- Signup voucher
- A $5 voucher is granted once cumulative recharges reach $5.
Strengths and limitations
Strengths
- Direct access to Kimi models from the company that trains them.
- Very low entry cost: a $1 recharge enables the API.
- Prepaid balance caps spend at what you have loaded.
Limitations
- No free tier; a recharge is required before any call.
- Tier 0 allows only 1 concurrent request and 3 requests per minute.
- Vouchers do not count toward the cumulative recharge that sets your tier.
- Only Moonshot's own models are available.
Who it suits
Good for
- Coding agents and tools built on Kimi models
- Low-budget experiments that can prepay a few dollars
Look elsewhere if
- Parallel or high-throughput workloads on a small balance
- Teams needing EU data residency
Alternatives to Kimi API (Moonshot AI Open Platform)
| Tool | From | Free option | Stages |
|---|---|---|---|
| Scaleway Generative APIs Inference provider (EU-hosted open-weight models) | Usage-based | ||
| Replicate Model hosting API (image, video, audio and LLMs) | $0.09/hr | ||
| OpenRouter Model router (unified API) | — | Free tier | |
| Hugging Face Inference Providers Router across inference providers | $9/mo | Free tier | |
| SambaNova Cloud (SambaCloud) Inference provider (open-weight models) | Usage-based | Free tier | |
| Novita AI Inference provider (open-weight models) | Usage-based |
Questions
How much does the Kimi K3 API cost?
On 2026-09-29 kimi-k3 cost $3 per 1M cache-miss input tokens, $0.30 per 1M cache-hit input tokens and $15 per 1M output tokens.
Is there a free Kimi API tier?
No. You must recharge at least $1 to start; a $5 voucher is added once your cumulative recharges reach $5.
How do I raise my Kimi API rate limit?
Limits rise with cumulative recharge: $10 gives 100 RPM and 15 concurrent requests, and further tiers start at $20, $100, $1,000 and $3,000.
Sources read for this page