AnthropicGA
Claude Opus 5
Frontier-class coding at the unchanged Opus price — half of Fable 5, and reporting higher numbers.
Context
1M
Max output
128K
Input /1M
$5.00
Output /1M
$25.00
Live pricing via OpenRouter
Best for
- Hard agentic coding and multi-file migrations
- Long-horizon autonomous work at $5/$25
- Teams that want the Fable-class ceiling without the Fable-class bill
Watch out
Benchmarks are vendor-reported. Published leaderboards disagree at the top of SWE-Bench Verified — DeepSeek V4 Pro 0813 reports 96.4% against this model’s 96.0% — so treat any ranking claim as source-dependent.
For creators. The judgment layer: review gates, ambiguous specs, and long-form work where the 1M window earns its keep.
Benchmarks
| swe bench verified | 96 |
| swe bench pro | 79.2 |
| frontier bench | 43.3 |
| arc agi 3 | 30.2 |
| gdpval aa | 1861 |
Capabilities
- 1M context, 128K max output
- SWE-Bench Verified 96.0% and Pro 79.2%, up from Opus 4.8's 88.6% / 69.2% (vendor-reported)
- Frontier-class agentic coding and computer use at unchanged Opus pricing