{
  "schemaVersion": 1,
  "modelId": "kimi-k3",
  "publicApiModelId": "kimi-k3",
  "availability": "available",
  "lastVerifiedAt": "2026-10-18T19:34:05.266405Z",
  "sourceUrls": [
    "https://platform.kimi.ai/docs/pricing/chat",
    "https://platform.kimi.ai/docs/guide/context-caching",
    "https://vllm.ai/blog/2026-10-16-kimi-k3-megakernel",
    "https://github.com/vllm-project/vllm/pull/61867"
  ],
  "pricingData": "/api/pricing.json",
  "headlineApiPriceChange": false,
  "specializedVllm": {
    "pullRequest": "https://github.com/vllm-project/vllm/pull/61867",
    "mergedAt": "2026-10-16T02:50:42Z",
    "mergeCommit": "642c18d64c4cb6db3ded8fbe1786718cd4005fcb",
    "selector": "decode_backend: latency_specialized",
    "default": "stock",
    "validatedGpu": "NVIDIA GB300",
    "tensorParallel": [
      8,
      16
    ],
    "maxDecodeBatch": 64,
    "interconnect": "one NVLink domain with multicast support",
    "mustPinContainingBuild": true
  },
  "labs": {
    "managedRouteIncluded": true,
    "route": "moonshotai/kimi-k3",
    "runId": "2026-10-18T04-18-21-026Z-cost-per-task-openrouter",
    "resultData": "/labs/cost-per-task.json",
    "hardwareReplayStatus": "blocked-no-authorized-gb300-hardware-harness",
    "requirements": [
      "authorized GB300 allocation and complete serving-cost ledger",
      "pinned vLLM, checkpoint, GPU/interconnect and precision settings",
      "same-hardware stock/optimized controls, real prompts, actual acceptance and prefill/decode latency",
      "repeated correctness, failures, energy and full operating-cost evidence"
    ],
    "noPaidInferenceOrHardwareRentalForThisRelease": true
  },
  "maintenance": {
    "owner": "aipricing-data",
    "cadenceDays": 1,
    "registryId": "kimi-k3-vllm-2026-10-18"
  }
}