Anthropic
Claude Opus 5 (Adaptive Reasoning, Low Effort)
Claude Opus 5 is the latest model in Anthropic's Opus family and a step-change improvement over Opus 4.8. It delivers major gains over Opus 4.8 in agentic coding, professional knowledge work, and long-horizon reasoning, and it is stronger per token across effort levels.
Introducing Claude OpusModel Snapshot
TypeLanguage
ReleaseJul 24, 2026
Context1,000,000
Tagstool-use, reasoning, vision, file-input
Overall
Rank #29 across 600
50.6
Coding
Rank #45 across 266
66.9
Blended Price
Rank #350 across 379
$10.00/M
Strength Profile
Percentile score by analysis domain.
* Cost is inverted: lower input, output, and blended prices rank higher.
Benchmark Percentiles
Higher bars mean stronger relative placement.
Model Details
Runtime
Output Speed54.8 tok/s
First Token3.24s
Max Output128,000
Pricing
Blended$10.00/M
Input$5.00/M
All Benchmarks
| Metric | Domain | Value | Rank |
|---|---|---|---|
| Artificial Analysis Intelligence Index | overall | 50.6 | #29 |
| Artificial Analysis Coding Index | coding | 66.9 | #45 |
| GPQA | reasoning | 88.9% | #55 |
| Humanity's Last Exam | reasoning | 41.3% | #17 |
| SciCode |