Private Kimi K3 API
Strongest open model. Limited capacity. Runs on Confidential Inference: an OpenAI-compatible endpoint inside TEEs. Fully verifiable.
How much does Kimi K3 cost on Confidential AI?
| Status | Input (per 1M tokens) | Input Cached (per 1M tokens) | Output (per 1M tokens) |
|---|---|---|---|
| On request | $3.00 | $0.30 | $15.00 |
Prices as of .
How do I call Kimi K3?
Kimi K3 is available on request. Request capacityand we'll email you when your access is live. Once it is, you call it through the same OpenAI-compatible API as every other model: get started.
Is Kimi K3 private on Confidential AI?
Yes. The model runs inside trusted execution environments: Intel TDX confidential VMs with NVIDIA GPUs in confidential computing mode. Prompts and responses stay invisible to the operator, Confidential AI included. Your client can fetch and verify the attestation before it sends anything. See attestation, or verify our production cluster in your browser.
How much slower is confidential inference?
About 1 to 3% lower token throughput on NVIDIA B200s for a correctly configured stack, with negligible impact on time to first token. Method and other measurements: performance overhead.
Full pricing: pricing.