{
 "study": "ai-tokenizer-inflation-2026",
 "date": "2026-10-28",
 "startedUtc": "2026-10-28T09:13:48.615Z",
 "methodSha256": "066cfe21c55a6db0ce68497f7f5a746a05fc6245bf1b1fedd2fe3c56b1a5fffa",
 "corpusManifestSha256": "0bd02f52f2b4e0b0e20992bfadae1592dc24d0cc9876b70ba2275f6b03a9310e",
 "items": [
  {
   "file": "anthropic-models.raw.json",
   "url": "https://api.anthropic.com/v1/models?limit=100",
   "retrievedUtc": "2026-10-28T09:13:49.490Z",
   "http": 200,
   "bytes": 13674,
   "sha256": "de9be779c4d815c911e9a6bc7480a1e868d893b864dcfeae6c729ec5159e9cf9"
  },
  {
   "file": "anthropic-count-tokens.raw.jsonl",
   "url": "https://api.anthropic.com/v1/messages/count_tokens",
   "retrievedUtc": "2026-10-28T09:14:12.282Z",
   "calls": 884,
   "failedCalls": 0,
   "bytes": 190990,
   "sha256": "7ae396c7abd74ed3a7b957142fe5bfddcfabd2d00f27d1d2988f69d91ee3526b"
  },
  {
   "file": "gemini-models.raw.json",
   "url": "https://generativelanguage.googleapis.com/v1beta/models?pageSize=200",
   "retrievedUtc": "2026-10-28T09:15:54.249Z",
   "http": 200,
   "bytes": 29174,
   "sha256": "49c2e20f2c3b5fc3539e387584565f6fb65142b628714df8b84487670d807ecc"
  },
  {
   "file": "gemini-countTokens.raw.jsonl",
   "url": "https://generativelanguage.googleapis.com/v1beta/models/<model>:countTokens",
   "retrievedUtc": "2026-10-28T09:17:00.336Z",
   "calls": 444,
   "failedCalls": 37,
   "quota429Retries": 0,
   "bytes": 156581,
   "sha256": "b4a285b5c7d94ffb54e3d8f32d0285031ac9d3d6de6e6406ac63c501d612ef7c"
  },
  {
   "file": "cohere-models.raw.json",
   "url": "https://api.cohere.com/v1/models",
   "retrievedUtc": "2026-10-28T09:17:00.522Z",
   "http": 200,
   "bytes": 5572,
   "sha256": "96819bd5a4e1a24cdf41fe0f26b4aabac9f9bf0fda99dcbdb8e3df1b422f28f2"
  },
  {
   "file": "cohere-tokenize.raw.jsonl",
   "url": "https://api.cohere.com/v1/tokenize",
   "retrievedUtc": "2026-10-28T09:18:00.175Z",
   "calls": 68,
   "failedCalls": 0,
   "note": "bodies are not stored in full (token arrays are large); each line keeps the response SHA-256 and length, and the count is the length of the tokens array",
   "bytes": 13380,
   "sha256": "2b0b64abe3abdaa0039d8c941d0ae1da5f27009c025719a7a683c01c7baa57d3"
  }
 ],
 "endedUtc": "2026-10-28T09:14:12.283Z",
 "docsChecked": {
  "url": "https://platform.claude.com/docs/en/build-with-claude/token-counting",
  "readUtc": "2026-10-28",
  "statements": [
   "Token counting is free to use but subject to requests per minute rate limits",
   "The token count is an estimate; counts might include tokens added automatically by Anthropic for system optimizations; you are not billed for system-added tokens",
   "Claude 4.7 and later models and Claude Mythos Preview use a newer tokenizer: approximately 30 percent more tokens for the same text, exact increase depends on content",
   "Fable and Mythos 5 and 5.1 share the tokenizer introduced with Claude Opus 4.7"
  ]
 }
}