Kimi K3 API Pricing
Moonshot · Frontier · Proprietary · released 2026-06. Prices in USD per 1 million tokens. Data snapshot: 2026-09-28.
The price we publish for Kimi K3 has changed 6 times since 2026-08-15, most recently on 2026-09-22 when the reference host changed (+76.5%); it is now the 2nd cheapest of 31 Frontier models. Buying from InferenceNet rather than at the rate quoted here lowers the token price by about 42%, for a quantized fp4 build. It is rated a stable release but does not meet all four criteria of the FinOps Friendly badge.
- Input: $3.00 per 1M tokens
- Output: $15.00 per 1M tokens
- Batch API (input/output): $2.28 / $11.40
- Prompt-cache read: $0.30
- Context window: 1M
- Arena ELO: 1488 (as of 2026-09-25)
- Available on: Direct API, OpenRouter (curated by hand, verify with the provider)
- AWS Bedrock regions: ap-northeast-1, ap-southeast-1, eu-central-1, eu-west-1, us-east-1, us-east-2, us-west-2
Kimi K3 market position
On blended price ($11.40 per 1M tokens at 30% input, 70% output) it is cheaper than 77% of Frontier models: 2nd cheapest of 31, tied with 6 others. Across every category it is cheaper than 8% of priced text models in the catalogue: 252nd cheapest of 281, tied with 6 others. Its Arena ELO of 1488 ranks 4th of the 68 models that carry one, tied with one other, above 94% of them.
Kimi K3 licence, size and capabilities
Kimi K3 is recorded here as proprietary: access is through an API, priced by the token. Its parameter count is not published. Listed capabilities: Text, Vision, Reasoning, Code, Agents. Filling its 1M context window once costs $3.00 in input tokens at list price.
Monthly cost at 100K requests, by business use case
From the after-discounts figure (prompt caching on the cached input share, batch API for async workloads, only where the provider publishes those rates), which assumes you implement both, to the list price. One figure means no discount is published.
| Use case | Tokens (in/out) | Monthly (after discounts – list) |
|---|---|---|
| Support Ticket | 1,500 / 500 | $957 – $1.2K |
| Knowledge Q&A | 2,000 / 800 | $1.4K – $1.8K |
| Meeting Summary | 10,000 / 1,200 | $3.4K – $4.8K |
| Marketing Content | 2,500 / 1,800 | $3.3K – $3.5K |
| Coding Task | 3,000 / 2,000 | $3.5K – $3.9K |
| Invoice Processing | 1,500 / 600 | $937 – $1.4K |
| Call Summary | 2,000 / 700 | $1.2K – $1.6K |
| Agent Workflow | 6,000 / 3,000 | $5.2K – $6.3K |
Kimi K3 pricing by host
A host is a company that runs the model and sells it by the token — the lab that made it, a cloud platform like Amazon Bedrock, Azure or Vertex, or an independent provider. One model, several sellers, and often several prices. This model is sold by 17 different hosts. The price quoted above is the one we publish so that models stay comparable; it is not necessarily the cheapest way to buy this model. List prices as published on 2026-09-28 via OpenRouter — negotiated rates, committed-use discounts and provisioned capacity are not shown. These hosts do not all serve the model at the same numeric precision, so the cheapest is not a like-for-like substitute for the dearest.
Cheapest host: InferenceNet at $1.00 input / $9.00 output per 1M tokens, 42% below the price we publish on a blended basis (30% input, 70% output). It serves a quantized fp4 build, which is not the same product as the full-precision model. The dearest, Alibaba, charges 2.0x the cheapest.
| Host | Input $/1M | Output $/1M |
|---|---|---|
| InferenceNet — fp4 | $1.00 | $9.00 |
| Wafer | $1.00 | $9.00 |
| Sail Research — fp4 | $1.03 | $9.04 |
| Morph — fp8 | $1.32 | $10.50 |
| Relace — fp4 | $2.10 | $10.50 |
| Makora | $2.04 | $12.75 |
| Phala | $2.55 | $12.75 |
| DigitalOcean | $2.55 | $12.95 |
| DeepInfra — bf16 | $2.85 | $14.25 |
| Chutes (the price shown above) — mxfp4 | $3.00 | $15.00 |
| Parasail (the price shown above) — fp4 | $3.00 | $15.00 |
| Modal (the price shown above) — mxfp4 | $3.00 | $15.00 |
| Together (the price shown above) | $3.00 | $15.00 |
| Fireworks (the price shown above) | $3.00 | $15.00 |
| BaseTen (the price shown above) — fp8 | $3.00 | $15.00 |
| Moonshot AI (the price shown above) — mxfp4 | $3.00 | $15.00 |
| Alibaba | $3.45 | $17.25 |
Kimi K3 price history
Kimi K3 has been in our daily price archive since tracking began on 2026-08-15, when it listed at $3.00 input / $15.00 output per 1M tokens. The price we publish has changed 6 times since then, 4 of them a change of reference host rather than a repricing. The latest change, on 2026-09-22, was +76.5% on the blended price (30% input, 70% output), to $3.00 input / $15.00 output (reference host changed, not a repricing). Net of all changes, the blended price is back where it started on 2026-08-15.
| Date | Old price (in / out) | New price (in / out) | Blended change |
|---|---|---|---|
| 2026-09-22 | $1.70 / $8.50 | $3.00 / $15.00 | +76.5% (reference host changed, not a repricing) |
| 2026-09-19 | $2.10 / $10.95 | $1.70 / $8.50 | -22.1% |
| 2026-09-18 | $3.00 / $15.00 | $2.10 / $10.95 | -27.2% (reference host changed, not a repricing) |
| 2026-09-17 | $2.65 / $13.28 | $3.00 / $15.00 | +13.0% (reference host changed, not a repricing) |
| 2026-09-13 | $2.34 / $11.70 | $2.65 / $13.28 | +13.5% |
1 earlier change is not listed.
Alternatives to Kimi K3
- Next cheaper Frontier model: GPT-5.4 (OpenAI) at $11.25 blended, 1% less.
- Next more expensive: Claude Opus 5.5 (Anthropic) at $15.20 blended, 33% more.
The nearest Frontier models on Arena ELO, each with its blended price and score difference against Kimi K3.
- Claude Opus 5 (Anthropic) — $5.00 in / $25.00 out per 1M, Arena ELO 1491: 67% more expensive, 3 ELO points higher
- GPT-5.6 Sol (OpenAI) — $4.00 in / $20.00 out per 1M, Arena ELO 1483: 33% more expensive, 5 ELO points lower
- Claude Opus 4.7 (Anthropic) — $5.00 in / $25.00 out per 1M, Arena ELO 1495: 67% more expensive, 7 ELO points higher
- Claude Opus 4.6 (Anthropic) — $5.00 in / $25.00 out per 1M, Arena ELO 1498: 67% more expensive, 10 ELO points higher
- Cheapest within 15 Arena points: GLM 5.3 Flash (Zhipu), 97% cheaper, 14 ELO points lower, out of 16 models in that band.
- Highest Arena ELO within 20% of this price: Kimi K3 itself has the highest score of the 6 models priced within 20% of it.
Kimi K3 head to head
Common questions about Kimi K3 pricing
- How much does Kimi K3 cost?
- Kimi K3 costs $3.00 per million input tokens and $15.00 per million output tokens on the API.
- Does Kimi K3 offer batch or cached pricing?
- Yes. Batch API pricing is $2.28/M input and $11.40/M output. Prompt-cache reads cost $0.30/M.
- What would Kimi K3 cost per month?
- For a support-ticket workload (1,500 input + 500 output tokens per request) at 100K requests/month, Kimi K3 costs about $1.2K at list prices.
- Has the price of Kimi K3 changed?
- Kimi K3 has been in our daily price archive since tracking began on 2026-08-15, when it listed at $3.00 input / $15.00 output per 1M tokens. The price we publish has changed 6 times since then, 4 of them a change of reference host rather than a repricing. The latest change, on 2026-09-22, was +76.5% on the blended price (30% input, 70% output), to $3.00 input / $15.00 output (reference host changed, not a repricing). Net of all changes, the blended price is back where it started on 2026-08-15.
These are list prices: real cost also depends on caching, batching, negotiated discounts, marketplace agreements, and existing contracts. Get a FinOps review from OptimNow or compare all models.
Build the business case for Kimi K3 in the AI ROI Calculator: add your business value on top of these costs to get ROI, break-even and payback.
All models · Compare side by side · Cloud compute pricing · Price barometer · FinOps guides · API and MCP server