{
  "updated_at": "2026-09-07T21:18:52.805842+00:00",
  "public_headline": "70B+",
  "description": "Token activity across cloud and local AI",
  "combined_approximate_tokens": 73594874003,
  "codex": {
    "lifetime_tokens": 61400000000,
    "basis": "Lifetime token counter reported from the Codex app dashboard by the owner on September 7, 2026. Includes input and output."
  },
  "claude": {
    "tokens": 2487324603,
    "basis": "Retained Claude Code usage records, January 28 through September 6, 2026, deduplicated by message and request. Includes input and output."
  },
  "local": {
    "estimated_generated_tokens": 9707549400,
    "gpus": 3,
    "gpu_model": "RTX 5090",
    "combined_tokens_per_second": 450,
    "start": "2026-01-01T00:00:00-05:00",
    "end": "2026-09-07T21:18:52.805842+00:00",
    "elapsed_seconds": 21572332,
    "basis": "Owner-supplied continuous throughput assumption for the three-GPU fleet combined. Estimated generated output; not a telemetry total.",
    "formula": "elapsed seconds \u00d7 450 combined tokens per second; GPU count is not multiplied again."
  },
  "notes": [
    "The headline is rounded down from the combined approximate total.",
    "Cloud traffic includes input and reused context; local throughput estimates generated output. This is a broad activity total, not a claim that all tokens are newly generated text.",
    "Hardware configuration and throughput are supplied by the owner. No utilization gaps or throughput variation are modeled."
  ]
}