Kimi K3 API pricing on Avian Compute
Every Avian Compute Kimi K3 rate we track, compared against 22 providers serving the same model.
The cheapest Kimi K3 input price we currently track is $0.192/1M on OpenRouter.
The cheapest Kimi K3 output price is $11.60/1M on HiAPI.
Avian Compute Kimi K3 rates
Prices per 1M tokens. Batch and cached-input rates are shown where the provider publishes them.
| Mode | Input | Output |
|---|---|---|
| Standard | $2.55/1M | $12.75/1M |
About Kimi K3
- Creator
- Moonshot AI
- Context
- 1.0M
- Modalities
- text, image
- Tool calling
- Yes
- Open weights
- Yes
Kimi K3 on other providers
- OpenRouter$0.192/1M
- IO.NET$1.98/1M
- EcoHash$2.00/1M
- HiAPI$2.32/1M
- IteraCompute$2.35/1M
- RouterPlus$2.50/1M
- PopNow$2.70/1M
- Deep Infra$2.85/1M
- Runcrate$2.99/1M
- Baseten$3.00/1M
- DigitalOcean$3.00/1M
- Empirio Labs$3.00/1M
- GMI Cloud$3.00/1M
- Lyceum$3.00/1M
- Novita AI$3.00/1M
- Perplexity$3.00/1M
- Runpod$3.00/1M
- Together AI$3.00/1M
- Prime Intellect$3.45/1M
- VoltageGPU$3.70/1M
- Heabsy$4.00/1M
About Avian Compute
Avian Compute provides AI inference and dedicated GPU compute services through NVIDIA B200 and B300 HGX nodes. It offers a range of AI models accessible via a single API key, billed on a pay-per-token basis or through long-term reservations.
Frequently Asked Questions
How much does Kimi K3 cost on Avian Compute?
Avian Compute serves Kimi K3 from $2.55 per 1M input tokens, across 1 tracked pricing mode. Prices are collected daily — see the table above for current input, output, and cached-input rates.
Is Avian Compute the cheapest way to run Kimi K3?
Avian Compute ranks #7 of 22 providers we track serving Kimi K3 on input price. The cheapest input price right now is on OpenRouter. The lowest output price is $11.60/1M on HiAPI. Rates change often, and throughput, latency and context limits differ between providers — check the comparison above before committing.
Does Avian Compute offer batch or cached pricing for Kimi K3?
We track 1 Kimi K3 offering from Avian Compute. Batch and cached-input rates appear in the table above when the provider publishes them; blank cells mean we have no current figure for that mode.