{
  "schemaVersion": 2,
  "verified": "2026-10-05",
  "activationCoverage": {
    "providers": 22,
    "tiers": 66,
    "supported": 6,
    "explicitly_unavailable": 3,
    "unestablished": 52,
    "blocked": 5
  },
  "observedNightlyAudit": {
    "verified": "2026-10-05",
    "providerRoutesScanned": 22,
    "tiersAccounted": 66,
    "priced": 42,
    "explicitlyUnavailable": 3,
    "unestablished": 16,
    "blocked": 5,
    "note": "The October 5 opening desk freshly re-audited all 22 provider routes and all 66 exact tracked models. Every activated route reproduced. Observed evidence remains 42 priced, 3 explicitly unavailable, 16 unestablished and 5 blocked; public Cache activation remains 6 supported, 3 explicitly unavailable, 52 unestablished and 5 blocked. GLM-5.3 publishes $0.26 cached input and limited-time free storage, but incomplete TTL, minimum, eligibility and composition terms keep public Cache calculation unestablished. No activation or observed status count changed."
  },
  "providerAudits": {
    "gemini": {
      "models": [
        "Flash-Lite 3.5",
        "Flash 3.6",
        "Pro 3.1"
      ],
      "sourceUrls": [
        "https://ai.google.dev/gemini-api/docs/pricing?hl=en",
        "https://ai.google.dev/gemini-api/docs/generate-content/caching?hl=en",
        "https://ai.google.dev/gemini-api/docs/models/gemini-3.1-pro-preview"
      ],
      "verified": "2026-10-05",
      "result": "Exact cached-input and storage prices were re-read. Flash 3.6 and Pro 3.1 also have exact current minimum-token rules and are activated. Flash-Lite 3.5 remains unestablished because the current caching guide does not list its exact minimum-token rule."
    },
    "openai": {
      "models": [
        "GPT-6 Luna",
        "GPT-5.6 Terra",
        "GPT-6 Sol"
      ],
      "sourceUrls": [
        "https://developers.openai.com/api/docs/models/gpt-6-luna",
        "https://developers.openai.com/api/docs/models/gpt-5.6-terra",
        "https://developers.openai.com/api/docs/models/gpt-6-sol",
        "https://developers.openai.com/api/docs/guides/prompt-caching"
      ],
      "verified": "2026-10-05",
      "result": "All three tracked OpenAI routes publish exact cached-input prices. GPT-6 Luna and GPT-6 Sol replaced GPT-5.6 Luna and GPT-5.6 Sol on 25 Sep 2026 by Will's decision: cache reads are 0.1x and writes 1.25x the uncached input rate, doubling above 272K input tokens. The prompt-caching guide establishes the 1,024-token minimum and 30-minute TTL. All three routes are activated."
    },
    "anthropic": {
      "models": [
        "Claude Haiku 4.5",
        "Claude Sonnet 5.5",
        "Claude Opus 5.5"
      ],
      "sourceUrls": [
        "https://platform.claude.com/docs/en/about-claude/pricing",
        "https://platform.claude.com/docs/en/models/sonnet-5-5/migration-guide"
      ],
      "verified": "2026-10-05",
      "result": "The nightly audit observes model-specific five-minute write, one-hour write and cache-read terms. Anthropic's 28 September first-party migration guide directs Sonnet 5 users to generally available Sonnet 5.5 at the same $2/$10 Standard and $1/$5 Batch basis; exact Cache read is $0.20, five-minute write $2.50 and one-hour write $4 per million tokens. Claude Opus 5.5 replaced Claude Fable 5.1 in the Flagship slot on 25 Sep 2026 by Will's decision. Exact public calculation remains disabled until the model-specific distinctions are promoted into this registry."
    },
    "deepseek": {
      "models": [
        "V4.1 Flash",
        "V4 Pro",
        "V4 Pro"
      ],
      "sourceUrls": [
        "https://api-docs.deepseek.com/quick_start/pricing/"
      ],
      "verified": "2026-10-05",
      "result": "The nightly audit observed V4.1 Flash cache hits at $0.006 peak and $0.003 off-peak per million tokens, plus exact V4 Pro peak and off-peak terms. Public calculation remains disabled until the time-window composition is promoted into this registry."
    },
    "mistral": {
      "models": [
        "Mistral Small 4",
        "Mistral Medium 3.5",
        "Mistral Large 3"
      ],
      "sourceUrls": [
        "https://docs.mistral.ai/inference/pricing"
      ],
      "verified": "2026-10-05",
      "result": "The nightly audit observed exact cached-input rows. Public calculation remains disabled until current exact rules are promoted into this registry."
    },
    "cohere": {
      "models": [
        "Command R7B (12-2024)",
        "Command R (08-2024)",
        "Command R+ (08-2024)"
      ],
      "sourceUrls": [
        "https://cohere.com/pricing",
        "https://docs.cohere.com/docs/models"
      ],
      "verified": "2026-10-05",
      "result": "The dated Command model table still marks all three exact SKUs Live. The current rendered Generative Models pricing tab reproduced Command R7B and Command R, and the current pricing FAQ reproduced Command R+ (08-2024) at $2.50/$10. No complete exact public cache-pricing rule was established."
    },
    "groq": {
      "models": [
        "GPT-OSS 20B",
        "GPT-OSS 120B",
        "GPT-OSS 120B"
      ],
      "sourceUrls": [
        "https://console.groq.com/docs/models",
        "https://console.groq.com/docs/batch"
      ],
      "verified": "2026-10-05",
      "result": "The nightly audit observed automatic prompt-cache pricing, inactivity TTL and Batch non-stacking rules. Public calculation remains disabled until all exact conditions are promoted into this registry."
    },
    "together": {
      "models": [
        "GPT-OSS 20B",
        "Llama 3.3 70B",
        "Kimi K2.6"
      ],
      "sourceUrls": [
        "https://www.together.ai/pricing",
        "https://docs.together.ai/docs/serverless/models",
        "https://www.together.ai/models/gpt-oss-20b",
        "https://www.together.ai/models/kimi-k26"
      ],
      "verified": "2026-10-05",
      "result": "Llama 3.3 70B remains reproducible in the serverless catalog. Two fresh October 1 watchdog reads of each exact GPT-OSS 20B and Kimi K2.6 model page reproduced $0.05/$0.20 and $1.20/$4.50, both exact pages published Serverless availability, and the Kimi page published $0.20 cached input. The distinct current serverless catalog still omits both models, so the conflict is preserved as a source reversal and both Cache routes remain blocked pending the independent close."
    },
    "perplexity": {
      "models": [
        "Sonar",
        "Sonar Reasoning Pro",
        "Sonar Pro"
      ],
      "sourceUrls": [
        "https://docs.perplexity.ai/docs/getting-started/pricing"
      ],
      "verified": "2026-10-05",
      "result": "No complete exact cache-pricing rule was established for the tracked Sonar routes."
    },
    "aws_bedrock": {
      "models": [
        "Claude Haiku 4.5 (Bedrock)",
        "Claude Sonnet 5 (Bedrock)",
        "Llama 4 Maverick 17B (Bedrock)"
      ],
      "sourceUrls": [
        "https://aws.amazon.com/bedrock/pricing"
      ],
      "verified": "2026-10-05",
      "result": "The October 1 closing watchdog rendered the current US East (Ohio) Anthropic and Meta panels and reproduced Claude Haiku 4.5 at $1/$5 and Llama 4 Maverick at $0.24/$0.97. Exact Claude Sonnet 5 remained absent, so only that Standard point stays carried. No complete exact Cache read, write, minimum, TTL, eligibility or composition rule was established; all three exact Cache routes remain unestablished."
    },
    "deepinfra": {
      "models": [
        "Llama 3.1 8B",
        "Llama 3.3 70B",
        "Gemma 4 31B Turbo"
      ],
      "sourceUrls": [
        "https://deepinfra.com/pricing",
        "https://api.deepinfra.com/models/list"
      ],
      "verified": "2026-10-05",
      "result": "DeepInfra's first-party API explicitly redirects the retired Maverick route to google/gemma-4-31B-it-turbo and publishes the successor at $0.09 input, $0.34 output and $0.05 cached input per million tokens. No complete exact write, storage, TTL, minimum, eligibility or Batch/promotion-composition rule was established, so observed evidence and public Cache activation remain unestablished."
    },
    "fireworks": {
      "models": [
        "GPT-OSS 20B",
        "GPT-OSS 120B",
        "DeepSeek V4 Pro"
      ],
      "sourceUrls": [
        "https://docs.fireworks.ai/serverless/pricing",
        "https://fireworks.ai/models/fireworks/gpt-oss-20b",
        "https://fireworks.ai/models/deepseek-ai/deepseek-v4-pro-0813"
      ],
      "verified": "2026-10-05",
      "result": "The exact current DeepSeek V4 Pro (0813) card reproduces $1.32 input, $0.044 cached input and $3.96 output for the tracked Pro route. The cached-input amount remains observed evidence only because complete write, storage, TTL, minimum, eligibility and composition terms are not established. GPT-OSS 20B remains blocked because its current model page says serverless is not supported and publishes no rate."
    },
    "cerebras": {
      "models": [
        "GPT-OSS 120B",
        "GPT-OSS 120B",
        "ZAI-GLM-4.7"
      ],
      "sourceUrls": [
        "https://api.cerebras.ai/public/v1/models"
      ],
      "verified": "2026-10-05",
      "result": "No complete exact cache rule was established for GPT-OSS 120B. ZAI-GLM-4.7 is blocked because it is absent from the current complete public catalog."
    },
    "sambanova": {
      "models": [
        "GPT-OSS 120B",
        "Llama 3.3 70B",
        "DeepSeek V3.1"
      ],
      "sourceUrls": [
        "https://cloud.sambanova.ai/plans/pricing"
      ],
      "verified": "2026-10-05",
      "result": "The official pricing table explicitly marks cached input unavailable for all three tracked routes."
    },
    "openrouter": {
      "models": [
        "Llama 3.3 70B",
        "DeepSeek V3",
        "DeepSeek R1"
      ],
      "sourceUrls": [
        "https://openrouter.ai/api/v1/models",
        "https://openrouter.ai/api/v1/models/deepseek/deepseek-chat-v3-0324/endpoints"
      ],
      "verified": "2026-10-05",
      "result": "Two fresh September 29 exact endpoint reads reproduced DeepSeek V3 0324 on SiliconFlow at $0.25/$1 with no cached-input rate and on GMICloud at $0.29/$1.14 with $0.11 cached input. The distinct current model-list API also confirmed the exact model. Observed evidence remains priced through GMICloud; public calculation remains unestablished because TokenScale's Standard row follows the cheapest single live route, which publishes no Cache rate."
    },
    "xai": {
      "models": [
        "Grok Build 0.1",
        "Grok 4.3",
        "Grok 4.5"
      ],
      "sourceUrls": [
        "https://docs.x.ai/developers/pricing"
      ],
      "verified": "2026-10-05",
      "result": "The nightly audit observed exact cached-input and long-context rows. Public calculation remains disabled until their composition rules are promoted into this registry."
    },
    "qwen": {
      "models": [
        "Qwen3.6-Flash",
        "Qwen3.7-Plus",
        "Qwen3.7-Max"
      ],
      "sourceUrls": [
        "https://www.alibabacloud.com/help/en/model-studio/model-pricing"
      ],
      "verified": "2026-10-05",
      "result": "The exact first-party QwenCloud model pages reproduced creation and hit prices for all three tracked models. The canonical Alibaba documentation route rendered only its application shell, so region-specific TTL, minimum and composition rules were not promoted. Public calculation remains disabled."
    },
    "glm": {
      "models": [
        "GLM-4.5-Air",
        "GLM-5",
        "GLM-5.3"
      ],
      "sourceUrls": [
        "https://docs.z.ai/guides/overview/pricing",
        "https://docs.z.ai/guides/overview/migrate-to-glm-new"
      ],
      "verified": "2026-10-05",
      "result": "The provider-directed GLM-5.3 successor publishes $0.26 cached input and limited-time free cache storage. Public calculation remains unestablished because complete TTL, minimum, eligibility and composition terms are not published."
    },
    "kimi": {
      "models": [
        "Kimi K2.5",
        "Kimi K2.6",
        "Kimi K3"
      ],
      "sourceUrls": [
        "https://platform.kimi.ai/docs/models",
        "https://platform.kimi.ai/docs/pricing/batch",
        "https://platform.kimi.ai/docs/pricing/chat-k26",
        "https://platform.kimi.ai/docs/pricing/chat-k3"
      ],
      "verified": "2026-10-05",
      "result": "The exact K2.6 and K3 pricing pages reproduced current Standard and cache-hit rates. K2.5 remains blocked because the current Model List says it is retired and calls return HTTP 404."
    },
    "minimax": {
      "models": [
        "MiniMax M2",
        "MiniMax M2.7",
        "MiniMax M3"
      ],
      "sourceUrls": [
        "https://platform.minimax.io/docs/guides/pricing-paygo",
        "https://platform.minimax.io/docs/api-reference/api-overview",
        "https://platform.minimax.io/docs/api-reference/anthropic-api-compatible-cache"
      ],
      "verified": "2026-10-05",
      "result": "Two fresh rendered pricing reads reproduced MiniMax M2 at $0.30 input, $1.20 output, $0.03 cache read and $0.375 cache write. The current API overview explicitly lists MiniMax-M2 as a supported Large Language Model API model, and the exact explicit-cache guide establishes cache_control with a five-minute lifetime. The Lite Cache route is activated."
    },
    "meta_llama": {
      "models": [
        "Muse Spark 1.2",
        "Muse Spark 1.2",
        "Muse Spark 1.2"
      ],
      "sourceUrls": [
        "https://ai.developer.meta.com/docs/pricing-rate-limits"
      ],
      "verified": "2026-10-05",
      "result": "A cache-busted rendered read reproduced Muse Spark 1.3, 1.2 and 1.1 at $0.15 cached input per million tokens alongside their Standard rows. Public calculation remains disabled until the exact tier eligibility, minimum and composition conditions are reproduced."
    },
    "sakana": {
      "models": [
        "Fugu Ultra",
        "Fugu Ultra",
        "Fugu Ultra"
      ],
      "sourceUrls": [
        "https://console.sakana.ai/pricing"
      ],
      "verified": "2026-10-05",
      "result": "The nightly audit observed cached-input rows through and above 272K. Public calculation remains disabled until the exact boundary rule is promoted into this registry."
    }
  },
  "rules": {
    "gemini": {
      "haiku": {
        "model": "Flash-Lite 3.5",
        "status": "unestablished",
        "note": "Google publishes a cached-input price and storage price, but the current caching guide does not publish the minimum-token rule for this exact tracked model. No estimate is made.",
        "source": "https://ai.google.dev/gemini-api/docs/pricing?hl=en",
        "verified": "2026-10-05"
      },
      "sonnet": {
        "model": "Flash 3.6",
        "status": "supported",
        "cacheSupported": true,
        "readRates": {
          "current": 0.075,
          "list": 0.15
        },
        "longContext": null,
        "write": {
          "status": "not_separately_published",
          "rates": {
            "current": null,
            "list": null
          },
          "longContextRates": null,
          "unit": null,
          "note": "Google does not publish a separate cache-write token rate for this route."
        },
        "storage": {
          "status": "priced",
          "rates": {
            "current": 0.5,
            "list": 1
          },
          "unit": "USD per 1M cached tokens per hour"
        },
        "minInputTokens": 4096,
        "minInclusive": true,
        "cacheType": "implicit_and_explicit",
        "defaultTtlSeconds": 3600,
        "combinesWithPromo": true,
        "eligibility": "Cache hits require at least 4,096 input tokens. Implicit caching is automatic but does not guarantee a hit; explicit caching guarantees the cached-token price and adds storage cost.",
        "source": "https://ai.google.dev/gemini-api/docs/pricing?hl=en",
        "eligibilitySource": "https://ai.google.dev/gemini-api/docs/generate-content/caching?hl=en",
        "verified": "2026-10-05"
      },
      "opus": {
        "model": "Pro 3.1",
        "status": "supported",
        "cacheSupported": true,
        "readRates": {
          "current": 0.2,
          "list": 0.2
        },
        "longContext": {
          "tokens": 200000,
          "inclusive": false,
          "readRates": {
            "current": 0.4,
            "list": 0.4
          }
        },
        "write": {
          "status": "not_separately_published",
          "rates": {
            "current": null,
            "list": null
          },
          "longContextRates": null,
          "unit": null,
          "note": "Google does not publish a separate cache-write token rate for this route."
        },
        "storage": {
          "status": "priced",
          "rates": {
            "current": 4.5,
            "list": 4.5
          },
          "unit": "USD per 1M cached tokens per hour"
        },
        "minInputTokens": 4096,
        "minInclusive": true,
        "cacheType": "implicit_and_explicit",
        "defaultTtlSeconds": 3600,
        "combinesWithPromo": false,
        "eligibility": "Cache hits require at least 4,096 input tokens. The cached-input price changes only when the input prompt exceeds 200,000 tokens.",
        "source": "https://ai.google.dev/gemini-api/docs/pricing?hl=en",
        "eligibilitySource": "https://ai.google.dev/gemini-api/docs/generate-content/caching?hl=en",
        "verified": "2026-10-05"
      }
    },
    "openai": {
      "haiku": {
        "model": "GPT-6 Luna",
        "status": "supported",
        "cacheSupported": true,
        "readRates": {
          "current": 0.01,
          "list": 0.01
        },
        "longContext": {
          "tokens": 272000,
          "inclusive": false,
          "readRates": {
            "current": 0.02,
            "list": 0.02
          }
        },
        "write": {
          "status": "priced",
          "rates": {
            "current": 0.125,
            "list": 0.125
          },
          "longContextRates": {
            "current": 0.25,
            "list": 0.25
          },
          "unit": "USD per 1M cache-write tokens",
          "note": "GPT-6 cache writes cost 1.25x the active uncached input-token rate."
        },
        "storage": {
          "status": "not_separately_published",
          "rates": {
            "current": null,
            "list": null
          },
          "unit": null,
          "note": "OpenAI publishes cache-read and cache-write token prices but no separate storage amount for this route."
        },
        "minInputTokens": 1024,
        "minInclusive": true,
        "cacheType": "implicit_and_explicit",
        "defaultTtlSeconds": 1800,
        "combinesWithPromo": false,
        "eligibility": "A visible prompt prefix must contain at least 1,024 input tokens. Cache reads cost 0.1x and writes cost 1.25x the active uncached input-token rate. Input prompts above 272,000 tokens use the published 2x input multiplier for the full request.",
        "source": "https://developers.openai.com/api/docs/models/gpt-6-luna",
        "eligibilitySource": "https://developers.openai.com/api/docs/guides/prompt-caching",
        "verified": "2026-10-05"
      },
      "sonnet": {
        "model": "GPT-5.6 Terra",
        "status": "supported",
        "cacheSupported": true,
        "readRates": {
          "current": 0.2,
          "list": 0.2
        },
        "longContext": {
          "tokens": 272000,
          "inclusive": false,
          "readRates": {
            "current": 0.4,
            "list": 0.4
          }
        },
        "write": {
          "status": "priced",
          "rates": {
            "current": 2.5,
            "list": 2.5
          },
          "longContextRates": {
            "current": 5,
            "list": 5
          },
          "unit": "USD per 1M cache-write tokens",
          "note": "GPT-5.6 cache writes cost 1.25x the active uncached input-token rate."
        },
        "storage": {
          "status": "not_separately_published",
          "rates": {
            "current": null,
            "list": null
          },
          "unit": null,
          "note": "OpenAI publishes cache-read and cache-write token prices but no separate storage amount for this route."
        },
        "minInputTokens": 1024,
        "minInclusive": true,
        "cacheType": "implicit_and_explicit",
        "defaultTtlSeconds": 1800,
        "combinesWithPromo": false,
        "eligibility": "A visible prompt prefix must contain at least 1,024 input tokens. Cache reads cost 0.1x and writes cost 1.25x the active uncached input-token rate. Input prompts above 272,000 tokens use the published 2x input multiplier for the full request.",
        "source": "https://developers.openai.com/api/docs/models/gpt-5.6-terra",
        "eligibilitySource": "https://developers.openai.com/api/docs/guides/prompt-caching",
        "verified": "2026-10-05"
      },
      "opus": {
        "model": "GPT-6 Sol",
        "status": "supported",
        "cacheSupported": true,
        "readRates": {
          "current": 0.2,
          "list": 0.2
        },
        "longContext": {
          "tokens": 272000,
          "inclusive": false,
          "readRates": {
            "current": 0.4,
            "list": 0.4
          }
        },
        "write": {
          "status": "priced",
          "rates": {
            "current": 2.5,
            "list": 2.5
          },
          "longContextRates": {
            "current": 5,
            "list": 5
          },
          "unit": "USD per 1M cache-write tokens",
          "note": "GPT-6 cache writes cost 1.25x the active uncached input-token rate."
        },
        "storage": {
          "status": "not_separately_published",
          "rates": {
            "current": null,
            "list": null
          },
          "unit": null,
          "note": "OpenAI publishes cache-read and cache-write token prices but no separate storage amount for this route."
        },
        "minInputTokens": 1024,
        "minInclusive": true,
        "cacheType": "implicit_and_explicit",
        "defaultTtlSeconds": 1800,
        "combinesWithPromo": false,
        "eligibility": "A visible prompt prefix must contain at least 1,024 input tokens. Cache reads cost 0.1x and writes cost 1.25x the active uncached input-token rate. Input prompts above 272,000 tokens use the published 2x input multiplier for the full request.",
        "source": "https://developers.openai.com/api/docs/models/gpt-6-sol",
        "eligibilitySource": "https://developers.openai.com/api/docs/guides/prompt-caching",
        "verified": "2026-10-05"
      }
    },
    "together": {
      "haiku": {
        "model": "GPT-OSS 20B",
        "status": "blocked",
        "note": "Two fresh exact model-page reads reproduce $0.05/$0.20, but the distinct current serverless catalog still omits the model. No current Cache rule can be claimed while the first-party surfaces disagree.",
        "source": "https://www.together.ai/pricing",
        "verified": "2026-10-05"
      },
      "opus": {
        "model": "Kimi K2.6",
        "status": "blocked",
        "note": "Two fresh exact model-page reads reproduce $1.20/$4.50, $0.20 cached input and serverless availability, but the distinct current serverless catalog still omits the model. No current Cache rule can be claimed while the first-party surfaces disagree.",
        "source": "https://www.together.ai/pricing",
        "verified": "2026-10-05"
      }
    },
    "fireworks": {
      "haiku": {
        "model": "GPT-OSS 20B",
        "status": "blocked",
        "note": "The exact current model page says serverless is not supported and the serverless pricing table exposes no current route.",
        "source": "https://fireworks.ai/models/fireworks/gpt-oss-20b",
        "verified": "2026-10-05"
      }
    },
    "cerebras": {
      "opus": {
        "model": "ZAI-GLM-4.7",
        "status": "blocked",
        "note": "The exact tracked model is absent from the current complete public catalog, so no current Cache rule can be claimed.",
        "source": "https://api.cerebras.ai/public/v1/models",
        "verified": "2026-10-05"
      }
    },
    "sambanova": {
      "haiku": {
        "model": "GPT-OSS 120B",
        "status": "explicitly_unavailable",
        "note": "The official pricing table marks cached input unavailable for this exact tracked route.",
        "source": "https://cloud.sambanova.ai/plans/pricing",
        "verified": "2026-10-05"
      },
      "sonnet": {
        "model": "Llama 3.3 70B",
        "status": "explicitly_unavailable",
        "note": "The official pricing table marks cached input unavailable for this exact tracked route.",
        "source": "https://cloud.sambanova.ai/plans/pricing",
        "verified": "2026-10-05"
      },
      "opus": {
        "model": "DeepSeek V3.1",
        "status": "explicitly_unavailable",
        "note": "The official pricing table marks cached input unavailable for this exact tracked route.",
        "source": "https://cloud.sambanova.ai/plans/pricing",
        "verified": "2026-10-05"
      }
    },
    "kimi": {
      "haiku": {
        "model": "Kimi K2.5",
        "status": "blocked",
        "note": "The current official Model List says this tracked model is retired and calls return HTTP 404, so no current Cache rule can be claimed.",
        "source": "https://platform.kimi.ai/docs/models",
        "verified": "2026-10-05"
      }
    },
    "minimax": {
      "haiku": {
        "model": "MiniMax M2",
        "status": "supported",
        "cacheSupported": true,
        "readRates": {
          "current": 0.03,
          "list": 0.03
        },
        "longContext": null,
        "write": {
          "status": "priced",
          "rates": {
            "current": 0.375,
            "list": 0.375
          },
          "longContextRates": null,
          "unit": "USD per 1M cache-write tokens",
          "note": "MiniMax publishes an exact cache-write token price for MiniMax M2."
        },
        "storage": {
          "status": "not_separately_published",
          "rates": {
            "current": null,
            "list": null
          },
          "unit": null,
          "note": "MiniMax publishes exact cache read and write token prices but no separate storage amount."
        },
        "minInputTokens": 0,
        "minInclusive": true,
        "cacheType": "explicit",
        "defaultTtlSeconds": 300,
        "combinesWithPromo": false,
        "eligibility": "MiniMax's Anthropic-compatible explicit prompt caching supports MiniMax M2 and M2-Stable with cache_control. Cached content has a five-minute lifetime that refreshes on a hit.",
        "source": "https://platform.minimax.io/docs/guides/pricing-paygo",
        "eligibilitySource": "https://platform.minimax.io/docs/api-reference/anthropic-api-compatible-cache",
        "verified": "2026-10-05"
      }
    }
  }
}