Moonshot AI › Kimi K3 › Pricing
$3 / $15per 1M tokens in / out
$0.30cache read · −90%
$2.00per task at max?
#20on OpenRouter by tokens · 1.7T / week
Price sheetper 1M tokens
| Input | $3 | list |
| Output | $15 | |
| Cache read | $0.30 | −90% vs input |
| Cache write · 5 min | $3 | 0% vs input |
| Context · max output | 1M · 944K | tokens |
Cost per task by effort?
low$1.1525
maxAPI default$2.0046.9
per taskHumanity's Last Exam
Same price per token at every effort; the bill grows with reasoning tokens. max costs 1× max for +0 points.
Where to buy21 providers · per 1M tokens
| Provider | Input | Output | Context |
|---|---|---|---|
| InferenceNet | $0.24 | $11.94 | 1M |
| Wafer | $0.36 | $11.90 | 1M |
| Makora | $0.65 | $15 | 1M |
| Relace | $0.80 | $3 | 1M |
| Sail Research | $0.90 | $6.75 | 1M |
| Morph | $1.13 | $14.90 | 1M |
| AkashML | $1.20 | $14 | 1M |
| Chutes | $1.85 | $9.50 | 1M |
| DigitalOcean | $2.55 | $12.95 | 1M |
| lab-qwen | $2.83 | $14.13 | — |
| DeepInfra | $2.85 | $14.25 | 1M |
| BaseTen | $3 | $15 | 1M |
| Featherless | $3 | $15 | — |
| Fireworks | $3 | $15 | 1M |
| lab-moonshotai | $3 | $15 | — |
| Modal | $3 | $15 | 1M |
| Moonshot AIlab | $3 | $15 | 1M |
| moonshotai | $3 | $15 | — |
| Novita | $3 | $15 | — |
| Together | $3 | $15 | 1M |
| Alibaba | $3.45 | $17.25 | 1M |
In Moonshot AI's line-upper 1M in / out
| Kimi K2.7 Code | $0.95 / $4 | −71% |
| Kimi K2.7 Code Highspeed | $1.90 / $8 | −43% |
| Kimi K3.1 Pro | $2 / $10 | −33% |
| Kimi K3 | $3 / $15 | this model |
| Kimi K3.1 Pro Highspeed | $4 / $20 | +33% |
Cheaper for the same workvs Kimi K3 at max
| Claude Opus 5.5 high | -9% | score +23% | $1.82 |
| GPT-6.1 Sol max | -64% | score +19% | $0.72 |
| GPT-6 Astra high | -14% | score +17% | $1.73 |
| Muse Spark 1.3 max | -20% | score +10% | $1.60 |
| GPT-6 Sol max | -48% | score +9% | $1.04 |
Pricing questions
How much does Kimi K3 cost?›
$3 per 1M input tokens and $15 per 1M output tokens. Cached input is $0.30. One Intelligence Index task costs about $2.00 at the default effort.
Is Kimi K3 free to use?›
There is no free API tier, but the weights are open: you can run it on your own hardware for the cost of compute.
How much does prompt caching save?›
Cache reads cost $0.30 per 1M, 90% below the input price. Writing to the cache costs $3 per 1M (5m TTL).
Is Kimi K3 cheaper than Kimi K2.6?›
No: $3 / $15 against $0.95 / $4, 250% more per token.
How does effort change the price?›
The price per token is the same at every effort. Higher effort spends more reasoning tokens: a task costs $1.15 at low, $2.00 at max.
What is a cheaper alternative to Kimi K3?›
Claude Opus 5.5 at high effort costs -9% per task with a score +23% on Humanity's Last Exam.