Claude Max 20×. Weekly allowance ~$2,155 API-equivalent.

Method

Divide corrected in-window API-equivalent consumption ($1,249.83) by the app-reported 58% of the weekly all-models quota. The window opened 2026-09-24 00:59:59 UTC and resets 2026-10-01 00:59:59 UTC. One record per message.id: input and cache counted once, output_tokens taken as the max across a message's rows. Costs are the harness's own api_equivalent_cost / api_list_price_equivalent_usd figures, not recomputed. Monthly allowance is that weekly figure ×4, a projection.

Corrected in-window consumption $1,249.83 ÷ 0.58 reported = $2,155 / week. Rounding band on the displayed 58% (57.5–58.5%): $2,136 – $2,174.

Monthly (× 4) is a projection of the measured weekly allowance, not a separately metered monthly quota. The report states ~$8,620, about 43× the $200 fee.

Higher confidence: the weekly allowance is measured by dividing corrected in-window consumption by the app-reported 58% quota. It is still one window, one percentage, and mix-dependent.

Measured cohort

Opus 5.5 grid, 18 tasks, all attempts. Harness api_equivalent_cost / api_list_price_equivalent_usd for 24 runs across 18 grid tasks, divided by the $2,155 weekly allowance. The report states $1,267.39 = 58.8% of one week.

The source report does not publish a grid-wide token total. Courtyard Blossom Stage 2's two attempts moved 1.49 billion tokens between them.

Corrected window consumption

Reported 58% used. Window opened 2026-09-24 00:59:59 UTC. Resets 2026-10-01 00:59:59 UTC.

Corrected window consumption $1,249.83. $1,249.83 ÷ 0.58 reported = $2,155 / week.

Per-task cost

Opus 5.5 rate card

Input / MTok

$4.00

Cache write, 5 minutes / MTok

$5.00

Cache write, 1 hour / MTok

$8.00

Cache read / MTok

$0.20

Output / MTok

$20.00

Solved by least squares from seven independent harness receipts, then confirmed against the card the repo records with verified_live_pricing: true. Write multipliers land on 1.25× and 2.00×. The card reproduces all seven receipts to 0.0000% error. Opus 5.5 is ~20% cheaper per token than Opus 5 ($5/$25, cache read $0.50).

Supporting observations

Remaining this week (42%)

~$905

Fee multiple

about 43× the $200 fee

Opus 5.5 benchmark work inside this window

$1,203.37, 96.3% of the $1,249.83 consumed, 55.8 of the reported 58 points

Prior-window Opus 5.5 benchmark work

$77.90 (the 09-22/23 robotics and CAD runs)

Grid at stated effort, dropping the Sekiro high retry

$1,256.22, or 58.3%

Courtyard Blossom Stage 2 share of the grid

59% of the grid's cost; the failed first attempt alone was $433.73

Eight robotics tasks

$30.46, 2.4% of the grid

Local transcripts versus harness receipts

Local transcripts capture $756 of the $1,281 of Opus 5.5 work, 59%. Local data alone would have put the allowance at $1,118.

Opus-5-dominated week of Aug 20–27

$4,101

Extra usage already spent

$72.41

Caveats

1. Mix dependence is the dominant term, and it is large. The quota meters weighted units, not dollars, so this $2,155 is denominated in an Opus-5.5-heavy week (96% of the window). Aligned to the same reset boundary, the Opus-5-dominated week of Aug 20–27 consumed $4,101 — nearly double this allowance, in a week that may not even have hit 100%. Opus 5.5 therefore burns roughly 2× more quota per API dollar than Opus 5. Quote $2,155 for Opus-5.5 work; expect closer to $4,000+ for Opus-5 work.
2. The plan string reads "Pro", not Max 20x. The app reports plan: "Pro". If that is literal rather than a coarse family label, this is a Pro allowance and the Max 20x figure would be higher. The same ambiguity appeared in the Codex audit. Worth confirming in billing.
3. Extra usage: $72.41 already spent. Overage is currently disabled but has been charged before, so some past weeks ran past their cap — meaning historical weekly totals are not clean ceiling observations.
4. Capture may still be incomplete. Five missing runs were recoverable because the harness had receipts. Non-benchmark Claude work outside local transcripts (claude.ai web, mobile, cloud) would be invisible and has no receipt. That pushes the allowance up, not down. Local transcripts capture only $756 of the $1,281 of Opus 5.5 work the harness has receipts for (59%). Using local data alone would have put the allowance at $1,118.
5. One window, one percentage. This is a single observation at one point in one window. It is far better evidence than the September 6 inference, but it is not a distribution.

Evidence

Record status: provisional.

• Source report
• Measurement JSON
• Manifest
• Checksums
• audit.py
• allow.py
• hist.py
• sess.py
• Published RemakeBench results