Model data profile
Data updated: 2026-08-06Claude Opus 5
Pricing, public benchmarks, source mapping and procurement checks in one evidence-led page.
Anthropic · US
Claude Opus 5
Data identifier: claude-opus-5
Official pricing sourcePublic benchmark matched1000K context
Blended price*$10USD / 1M tokens
API pricing and context
Last observed: 2026-08-06. Vendor pages remain authoritative.
* Blended price uses 75% input and 25% output for consistent comparison. It does not represent a real workload. Basis: Official API list price.
API cost estimator
A mechanical estimate using current input and output list prices. Cache, batch discounts, retries and taxes are excluded.
+=
Estimated cost—USD
Public third-party evidence
TANDIU does not run or reproduce these tests. Results are shown with their public source and model mapping.
Artificial Analysis
Intelligence index61
Output speed54 tok/s
Time to first chunk62.83 sec
Source mappingVariant match
Model name in sourceClaude Opus 5 (max)Open source leaderboard ↗
LiveBench
Overall score80.09
Reasoning91.21
Coding81.45
Agentic Coding65.2
Source mappingVariant match
Model name in sourceclaude-opus-5-max-effortOpen source leaderboard ↗
Evidence boundary
Observed facts versus items to validate
- Pricing and context
- Recorded from the vendor pricing source; observed 2026-08-06
- Third-party metrics
- Artificial Analysis (Variant match); LiveBench (variant match)
- Business suitability
- No unverified industry-performance claim is made; validate with your own tasks
Procurement checks
Confirm before a pilot
- API availability, payment and invoicing in the target region
- Data retention, training policy, hosting region and compliance
- Structured output, tool use, multimodality and version stability
- Task success rate, latency, rate limits and retry cost on real workloads
Continue comparing
Do not rely on one score
Aggregate metrics can narrow a shortlist. Final comparisons should use the same business inputs, parameters and success criteria.
Claude Opus 5 FAQ
Short answers about the data and how to use this page.
How is the blended price calculated?
It uses 75% input price plus 25% output price as a consistent comparison basis. It is not an estimate of any specific company's bill.
Does TANDIU run the intelligence benchmark?
No. Metrics come from the labeled public third-party source. TANDIU collects the data, maps model versions and links back to the source.
Can this page determine a procurement decision?
No. Pricing and public metrics support shortlisting only. Contracts, region, compliance, features and task-level acceptance tests still need review.