{
  "schemaVersion": 1,
  "title": "Prompt cache break-even evidence set",
  "sourceCheckedAt": "2026-08-22",
  "currency": "USD",
  "rateUnit": "per_1m_tokens",
  "publicationStatus": "article_approved",
  "articleApproval": {
    "status": "approved",
    "approvedAt": "2026-08-22T23:03:18+08:00",
    "approvedBy": "xiaofengzhang"
  },
  "scope": "Simulated cost of one million reusable prefix tokens. Output and changing suffix tokens are excluded because they are identical between the cached and uncached alternatives.",
  "formulas": {
    "writePricedCache": {
      "uncached": "K * B",
      "cached": "W + (K - 1) * R",
      "firstProfitableTotalUses": "smallest integer K where K * B > W + (K - 1) * R"
    },
    "storagePricedCache": {
      "uncached": "N * B",
      "cached": "S * T + N * R",
      "firstProfitableSubsequentReads": "smallest integer N where N * B > S * T + N * R"
    },
    "bestEffortCache": {
      "expectedCost": "Q * (H * R + (1 - H) * B)",
      "expectedSavings": "Q * H * (B - R)"
    }
  },
  "providers": [
    {
      "provider": "OpenAI",
      "modelId": "gpt-5-6-sol",
      "model": "GPT-5.6 Sol",
      "mechanism": "explicit_or_implicit_prefix_cache",
      "minimumPrefixTokens": 1024,
      "ttl": "30m",
      "baseInputRate": 5,
      "cacheWriteRate": 6.25,
      "cacheReadRate": 0.5,
      "storageRatePerHour": null,
      "deterministicBreakEven": true,
      "sourceUrls": [
        "https://developers.openai.com/api/docs/guides/prompt-caching",
        "https://platform.openai.com/pricing"
      ]
    },
    {
      "provider": "Anthropic",
      "modelId": "claude-sonnet-5",
      "model": "Claude Sonnet 5",
      "mechanism": "automatic_or_explicit_prefix_cache",
      "minimumPrefixTokens": 1024,
      "ttl": "5m_or_1h",
      "baseInputRate": 2,
      "cacheWrite5mRate": 2.5,
      "cacheWrite1hRate": 4,
      "cacheReadRate": 0.2,
      "storageRatePerHour": null,
      "deterministicBreakEven": true,
      "sourceUrls": [
        "https://platform.claude.com/docs/en/build-with-claude/prompt-caching",
        "https://platform.claude.com/docs/en/about-claude/pricing"
      ]
    },
    {
      "provider": "Google",
      "modelId": "gemini-3-7-flash",
      "model": "Gemini 3.7 Flash",
      "mechanism": "explicit_cache_in_generate_content_api",
      "minimumPrefixTokens": 4096,
      "ttl": "configurable_default_1h",
      "baseInputRate": 0.75,
      "cacheWriteRate": null,
      "cacheReadRate": 0.075,
      "storageRatePerHour": 0.5,
      "deterministicBreakEven": true,
      "rateValidThrough": "2026-12-31",
      "billingAssumption": "The explicit-cache comparison includes cache storage plus subsequent cached-token reads. The reviewed Google documentation does not enumerate a separate cache-write token rate. Verify the actual invoice export before forecasting a production budget.",
      "sourceUrls": [
        "https://ai.google.dev/gemini-api/docs/caching?hl=en",
        "https://ai.google.dev/gemini-api/docs/generate-content/caching?hl=en",
        "https://ai.google.dev/gemini-api/docs/pricing?hl=en"
      ]
    },
    {
      "provider": "DeepSeek",
      "modelId": "deepseek-v4-flash",
      "model": "DeepSeek V4 Flash",
      "mechanism": "automatic_best_effort_prefix_cache",
      "minimumPrefixTokens": null,
      "minimumPrefixTokensStatus": "not_confirmed_in_reviewed_source",
      "ttl": "provider_managed_hours_to_days",
      "baseInputRate": 0.44,
      "cacheWriteRate": null,
      "cacheReadRate": 0.014,
      "storageRatePerHour": null,
      "deterministicBreakEven": false,
      "sourceUrls": [
        "https://api-docs.deepseek.com/guides/kv_cache/",
        "https://api-docs.deepseek.com/quick_start/pricing/"
      ]
    }
  ],
  "breakEvenCases": [
    {
      "id": "openai-sol-30m",
      "scenarioType": "write_priced",
      "modelId": "gpt-5-6-sol",
      "prefixTokens": 1000000,
      "ttlHours": 0.5,
      "firstProfitableTotalUses": 2,
      "cacheHitsAtBreakEven": 1,
      "uncachedCost": 10,
      "cachedCost": 6.75,
      "savings": 3.25
    },
    {
      "id": "claude-sonnet-5-5m",
      "scenarioType": "write_priced",
      "modelId": "claude-sonnet-5",
      "prefixTokens": 1000000,
      "ttlHours": 0.08333333333333333,
      "firstProfitableTotalUses": 2,
      "cacheHitsAtBreakEven": 1,
      "uncachedCost": 4,
      "cachedCost": 2.7,
      "savings": 1.3
    },
    {
      "id": "claude-sonnet-5-1h",
      "scenarioType": "write_priced",
      "modelId": "claude-sonnet-5",
      "prefixTokens": 1000000,
      "ttlHours": 1,
      "firstProfitableTotalUses": 3,
      "cacheHitsAtBreakEven": 2,
      "uncachedCost": 6,
      "cachedCost": 4.4,
      "savings": 1.6
    },
    {
      "id": "gemini-3-7-1h",
      "scenarioType": "storage_priced",
      "modelId": "gemini-3-7-flash",
      "prefixTokens": 1000000,
      "ttlHours": 1,
      "firstProfitableSubsequentReads": 1,
      "uncachedCost": 0.75,
      "cachedCost": 0.575,
      "savings": 0.175
    },
    {
      "id": "gemini-3-7-4h",
      "scenarioType": "storage_priced",
      "modelId": "gemini-3-7-flash",
      "prefixTokens": 1000000,
      "ttlHours": 4,
      "firstProfitableSubsequentReads": 3,
      "uncachedCost": 2.25,
      "cachedCost": 2.225,
      "savings": 0.025
    },
    {
      "id": "gemini-3-7-24h",
      "scenarioType": "storage_priced",
      "modelId": "gemini-3-7-flash",
      "prefixTokens": 1000000,
      "ttlHours": 24,
      "firstProfitableSubsequentReads": 18,
      "uncachedCost": 13.5,
      "cachedCost": 13.35,
      "savings": 0.15
    }
  ],
  "deepSeekHitRateCases": [
    { "requests": 100, "prefixTokens": 1000000, "hitRate": 0, "expectedCost": 44, "savings": 0 },
    { "requests": 100, "prefixTokens": 1000000, "hitRate": 0.25, "expectedCost": 33.35, "savings": 10.65 },
    { "requests": 100, "prefixTokens": 1000000, "hitRate": 0.5, "expectedCost": 22.7, "savings": 21.3 },
    { "requests": 100, "prefixTokens": 1000000, "hitRate": 0.75, "expectedCost": 12.05, "savings": 31.95 },
    { "requests": 100, "prefixTokens": 1000000, "hitRate": 1, "expectedCost": 1.4, "savings": 42.6 }
  ]
}