Model Details
Where is Claude Opus 5 a perfect fit?
- Difficult, repository-scale coding, large refactors, and end-to-end software development
- Long-running autonomous agents that must plan, execute, use tools, and recover from mistakes
- High-precision code review, debugging, and bug discovery
- Deep reasoning and complex multi-stage professional tasks
- Enterprise knowledge work involving documents, spreadsheets, and presentations
- Long-context analysis across very large codebases, documents, and datasets
- Vision-heavy workflows involving charts, diagrams, documents, and UI understanding
- Multi-agent workflows requiring coordination between specialized subagents
Quick Model Estimate
Your Claude Opus 5 Cost Estimate
💰 Total Cost
—
for 1000 input + 1000 output tokens
Cost Breakdown
Prices are watched for changes and checked against the provider's own page.
Pricing
|
Provider
↕
|
Modality
↕
|
Service Tier
↕
|
Input Price
(per 1M tokens) ↕ |
Output Price
(per 1M tokens) ↕ |
Cached Input
(per 1M tokens) ↕ |
Cached Output
(per 1M tokens) ↕ |
Context Size
↕
|
View
|
|---|---|---|---|---|---|---|---|---|
 Anthropic | Text | Standard | $5.0000 | $25.0000 | $0.5000 | $10.0000 | 1,000,000 tokens | → |
| Anthropic | Text | Batch | $2.5000 | $12.5000 | N/A | N/A | 1,000,000 tokens | → |
 Perplexity | Text | Standard | $5.0000 | $25.0000 | $0.5000 | N/A | 1,000,000 tokens | → |
Benchmarks
Scores from standardized evaluations by Artificial Analysis and Design Arena. Higher is better — the indices summarize overall ability, while the detailed scores break down performance on individual benchmarks.
Artificial Analysis
Higher is better · benchmarked by Artificial AnalysisGraduate-level questions in biology, chemistry & physics
Expert-level questions across many academic domains
Unpublished physics research reasoning problems
Long-context reasoning across large inputs
Research-level scientific coding tasks
Economically valuable, real-world knowledge work
Breadth of factual knowledge across domains
How reliably the model avoids fabricated answers
| Variant | GPQA Diamond | Humanity's Last Exam | AA-LCR | SciCode | CritPt | GDPval | Omniscience Accuracy | Omniscience Non-Hallucination |
|---|---|---|---|---|---|---|---|---|
| Claude Opus 5 (Adaptive Reasoning, Max Effort) | 93.2% | 54.9% | 79.3% | 56.4% | 29.1% | 60.4% | 60.9% | 39.2% |
| Claude Opus 5 (Max) | 93.2% | 54.9% | 79.3% | 56.4% | 29.1% | 61.1% | 60.9% | 39.2% |
| Claude Opus 5 (Adaptive Reasoning, Xhigh Effort) | 93.7% | 54.4% | 80.3% | 55.7% | 27.7% | 58.8% | 59.5% | 40.5% |
| Claude Opus 5 (Xhigh) | 93.7% | 54.4% | 80.3% | 55.7% | 27.7% | 59.6% | 59.5% | 40.5% |
| Claude Opus 5 (Adaptive Reasoning, High Effort) | 93.7% | 52.8% | 79% | 55.4% | 28.3% | 54% | 58.9% | 38.8% |
| Claude Opus 5 (High) | 93.7% | 52.8% | 79% | 55.4% | 28.3% | 54.7% | 58.9% | 38.8% |
| Claude Opus 5 (Adaptive Reasoning, Medium Effort) | 91.9% | 51.3% | 82% | 51.5% | 26.9% | 48.8% | 57.1% | 39.3% |
| Claude Opus 5 (Medium) | 91.9% | 51.3% | 82% | 51.5% | 26.9% | 49.5% | 57.1% | 39.3% |
| Claude Opus 5 (Adaptive Reasoning, Low Effort) | 88.9% | 43.4% | 81.3% | 49.2% | 23.1% | 39.7% | 55.9% | 37.8% |
| Claude Opus 5 (Low) | 88.9% | 43.4% | 81.3% | 49.2% | 23.1% | 40.1% | 55.9% | 37.8% |
Design Arena
Elo rating by arena & category| Variant | Arena | Category | Elo | Win % | Percentile | Avg time (ms) |
|---|---|---|---|---|---|---|
| claude-opus-5 | agents | agenticgamedev | 1,261 | 53.8% | 92 | — |
| claude-opus-5 | agents | androidnative | 1,269 | 55.9% | 92 | — |
| claude-opus-5 | agents | fullstack | 1,306 | 57.3% | 91 | — |
| claude-opus-5 | agents | godotgamedev | 1,279 | 54.2% | 92 | — |
| claude-opus-5 | agents | mobileapps | 1,348 | 67.1% | 98 | — |
| claude-opus-5-xhigh | agents | mobileapps | 1,297 | 60% | 93 | — |
| claude-opus-5 | agents | python-pptxslides | 1,267 | 54.4% | 85 | — |
| claude-opus-5 | agents | webapps | 1,260 | 54.5% | 81 | — |
| claude-opus-5 | models | 3d | 1,335 | 59.6% | 91 | 331,260 |
| claude-opus-5 | models | asciiart | 1,347 | 66.5% | 99 | 245,574 |
| claude-opus-5 | models | codecategories | 1,326 | 57.3% | 94 | 315,071 |
| claude-opus-5 | models | dataviz | 1,331 | 58.3% | 94 | 223,259 |
| claude-opus-5 | models | gamedev | 1,342 | 58.5% | 95 | 392,812 |
| claude-opus-5 | models | svg | 1,312 | 58.4% | 97 | 105,526 |
| claude-opus-5 | models | uicomponent | 1,349 | 60.6% | 96 | 221,527 |
| claude-opus-5 | models | website | 1,311 | 56% | 94 | 311,143 |
Compare Claude Opus 5 with
Two models side by side: worked costs, every rate, and every benchmark figure both of them report.
Other Models in the Claude 5 Family
FAQs about Claude Opus 5
How much does Claude Opus 5 cost per 1M tokens?
Claude Opus 5 costs between $2.50 and $5.00 per million input tokens, and between $12.50 and $25.00 per million output tokens. The range covers Standard and Batch pricing.
What does a typical workload cost with Claude Opus 5?
1,000 requests of 2,000 input and 500 output tokens each — 2,000,000 input and 500,000 output tokens in total — costs $11.25 with Claude Opus 5 at its lowest rates. Use the calculator on this page for your own volumes.
Which provider is cheapest for Claude Opus 5?
Anthropic has the lowest input rate for Claude Opus 5, at $2.50 per million tokens. Across all providers: Anthropic from $2.50 and Perplexity from $5.00 per million input tokens.
Is there cheaper Claude Opus 5 pricing than the standard rate?
Yes. Batch pricing for Claude Opus 5 starts at $2.50 per million input tokens against $5.00 on the standard tier, about 50% lower.
What is Claude Opus 5's context window?
Claude Opus 5 has a context window of 1,000,000 tokens, and returns up to 128,000 tokens in a single response.
What input types does Claude Opus 5 support?
Claude Opus 5 accepts text, image and file input, and returns text.