As checked 2026-10-25: Anthropic’s 12-page PDF expressly dates its list prices May 27, 2026, identifies the Claude API as the pricing platform, and enumerates model, inference scope, context bracket, standard or batch tier and several cache rates. This is a dated economics guide, not a permanent price guarantee or a personal hands-on benchmark.
Scope: global English reference; USD list prices where shown; local eligibility, currency, taxes, negotiated terms and model release dates require separate confirmation. Dynamic source-page prices have a check date, not a verified initial effective date.
Why this source is unusually useful
Anthropic’s 12-page PDF expressly dates its list prices May 27, 2026, identifies the Claude API as the pricing platform, and enumerates model, inference scope, context bracket, standard or batch tier and several cache rates. That structure is far stronger for a historical archive than an undated pricing widget. The rates in this article therefore identify the effective date from the document, while SKU availability after that date still requires separate verification. A PDF list price is not a private enterprise quote or a universal guarantee of geographic access.
A representative standard-rate comparison
For global Claude API list prices in the dated PDF, Claude Opus 4.8 is shown at $5 per million base input tokens and $25 per million output tokens, with $6.25 for five-minute cache writes, $10 for one-hour writes and $0.50 for cache hits. Claude Sonnet 4.6 is shown at $3 input and $15 output, with cache rates of $3.75, $6 and $0.30. Claude Haiku 4.5 is listed at $1 input and $5 output. These prices are not interchangeable across model versions or processing arrangements.
The geographic branch matters
The PDF lists both global and US-only inference prices for selected later Claude 4.x models. For Claude Opus 4.8, its US-only standard listing is $5.50 input and $27.50 output, rather than the global rates above. The change is a commercially separate offer and must be represented by its own region and tier values. Other models and platforms do not necessarily share the same coverage. Cloud partner endpoints and contract-specific residency terms should be checked independently.
How to estimate cache-aware batch usage
Suppose a batch of independent records generates 5 million base input tokens and 1 million output tokens on Claude Sonnet 4.6. The dated global batch list rate of $1.50 input and $7.50 output per million gives $7.50 + $7.50 = $15 in the simple no-cache scenario. Add cache-write or cache-hit rates separately if the request actually uses them, and do not claim the service will process all jobs by a user-defined deadline without checking its current batch restrictions.
Choosing what to trust
A comparison should preserve model family and number, global versus US-only scope, cache lifetime, batch eligibility and context bracket. A consumer Claude subscription advertised on a different page is not part of this API list price. The official live Claude platform documentation can be used for rechecks, but it should not overwrite the dated source or create a fictitious May-effective price for a later model.
What the dated PDF does not establish
The document is a published vendor list tariff, not a dated transaction log for an arbitrary organization. It cannot prove which model a customer selected, whether a SKU was offered in a particular country at a later cutoff, how a private agreement changed rates or whether a request qualified for cache hits. The specific price-effective date is meaningful for the public list at release, but no unverified end date should be fabricated. An archive should preserve the PDF alongside any later notice superseding parts of it.
A cache arithmetic check
Assume a Claude Sonnet 4.6 global standard request writes two million tokens into a five-minute cache once and subsequently hits that identical material four times. Using the dated PDF’s listed cache-write $3.75 and hit $0.30 per million rates, that portion is 2 × $3.75 plus 4 × 2 × $0.30, or $9.90, excluding any other input/output tokens and applicable extras. The example assumes all hits are eligible and that retention is maintained; the platform’s actual cache-control semantics determine whether those assumptions hold.
Decision checklist
- Identify the specific plan or API endpoint, model version, region and currency before pricing a job.
- Measure input, output, caching, generated media and failed-job units separately.
- Check the official pricing page again before procurement, especially if the SKU has no effective-date notice.
- Keep source evidence and a timestamped price ID; do not treat unknowns as zero.
Pricing evidence boundary
This cost discussion concerns Anthropic. Reproduce a quote only with its exact model or plan identifier, chargeable unit, billing region, processing mode and timestamp; a changing vendor rate card is not a permanent price guarantee. If any one of these dimensions is missing, keep the associated rate and worked scenario under editorial review instead of filling the gap with an estimate.
Sources & Further Reading
- Official list-price PDF with effective date 2026-05-27 and scoped Claude API rate rows — official first-party source, accessed 2026-10-25; official price-effective date 2026-05-27.
- Live Claude developer pricing explains model, caching, batch and partner channels — official first-party source, accessed 2026-10-25; original effective date not specified in source.
- Claude individual monthly subscription and team purchasing are distinct from API — official first-party source, accessed 2026-10-25; original effective date not specified in source.