[xujustinj]

Argon's Pro costs used introductory pricing. Reprice its saved usage at standard rates ($4/M input, $20/M output including thinking, $0.20/M cached input), leaving helper-model rates unchanged. Overall cost becomes $0.361595/page, previously $0.181650/page; all six challenge costs are updated too. Scores are unchanged.

The same 100 cost samples reconcile to $329.3938376: $327.1692126 Argon + $2.224625 helpers. This preserves the existing 93 original / seven replacement measurements and the assumption of no additional charges for unreported usage; it is a benchmark estimate, not historical billing. No inference was rerun, and no raw traces or document-level artifacts are included.

The public CLI calculator now uses standard rates consistently, and the README states the pricing policy.

Test plan:

• Reconciled all 100 previous costs to the published ledger, then independently checked their repriced usage against the public calculator.
• Verified exactly seven CSV cost cells changed; every score and other provider is unchanged.
• Pricing regression red/green: the existing native-usage test fails with introductory rates and passes after the fix. All 25 offline Gemini provider/transport tests pass; no new tests added.
• README regeneration/check, focused Ruff, and diff checks pass.

Pricing source: https://ai.google.dev/gemini-api/docs/latest-model#pricing