[BUG] Admin cost_report API reporting ~100× above token-rate math — $3,147 vs expected $42 for the past 7 days

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

Hi — I'm an individual API customer. I saw the issue on the console, and had Claude Code assist in investigation.

Requests are given below.

The admin /v1/organizations/cost_report endpoint reports numbers exactly ~100× higher than what published per-token rates predict. Sampled days over the past week:

| day (UTC) | cost_report amount | token × published rate |
|---|---|---|
| 2026-04-17 | $845.72 | $8.44 |
| 2026-04-18 | $1,100.68 | ~$11 |
| 2026-04-19 | $331.57 | ~$3.30 |

7-day total reported: $3,147. Computed-from-tokens at published Claude Sonnet 4 rates ($3/M input, $15/M output, $0.30/M cache-read) across 9.86M input / 0.75M output / 4.41M cache_read: ~$42.

Every line item in the per-description breakdown is exactly ~100× too high against its own token_type / context_window / service_tier count.

Console balance currently shows $21.40 (auto-reload disabled). Application-side logs, which compute cost locally via the published-rate formula, total $17.55 for the week — matching my expected ~$42 divided by 100× about exactly.

Sample request IDs (several, across days):

• req_011CaC5oUBw5qhgzauytoiL2
• req_011CaC67j5sz8rxYnJ53x9oP
• req_011CaC6QTVVU3X4u7fcLpAS9

Thanks

What Should Happen?

Two possibilities:

1. The cost_report amount field is returned in a sub-unit of USD (cents / hundredths) despite the "currency": "USD" label → documentation / API bug, my real spend is ~$31 for the week, no refund needed, but please confirm and update the docs.
2. My account will actually get billed ~$3,100 at ~100× list rate → I'd like a refund and an explanation of the upstream bug.

Either way I need a human to confirm which interpretation applies.

Error Messages/Logs

Steps to Reproduce

Look at the console usage report for an api account, and see scary numbers.

Claude Model

Sonnet (default)

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.114 (Claude Code)

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

_No response_