Anthropic
Claude Opus 4.8
Release date: 5/28/2026
A long-context reasoning model optimized for complex professional and agentic workflows.
Pramaana IndexAccuracy: StrongLatency: StandardCost: Strong
- Accuracy / Pramaana Index
- 69.90%± 0.82
- Latency / Pramaana Index
- 904.74s
- Cost / Test / Pramaana Index
- $2.10
- Context window
- 500k
- Max output tokens
- 96k
- Input modality
- Text · Image · File
Hyperparameter settings
temperature: 0.0
reasoning: extended
fallbacks: disabled
Benchmark performance for Claude Opus 4.8
| Benchmarks | Accuracy | Ranking |
|---|---|---|
| Pramaana Index | 69.90% ± 0.82 | 3 / 35 |
| Pramaana Multimodal Index | 65.82% ± 0.99 | 3 / 24 |
| Legal Research Fidelity | 70.10% ± 1.16 | 2 / 18 |
| Financial Modeling Agent | 60.01% ± 1.33 | 3 / 22 |
| Clinical Documentation Safety | 63.60% ± 1.50 | 3 / 19 |
| MATH-500 | 77.70% ± 1.67 | 5 / 112 |
| GPQA Diamond | 81.71% ± 0.48 | 4 / 121 |
| MMLU-Pro | 79.20% ± 0.65 | 4 / 119 |
| Adaptive Tutoring Dialogue | 65.50% ± 0.82 | 4 / 16 |
| SWE-bench Verified | 60.01% ± 0.99 | 3 / 69 |
| LiveCodeBench | 76.92% ± 1.16 | 3 / 127 |
| Agentic Reliability | 54.02% ± 1.33 | 3 / 12 |
| Strategic Games Arena | 62.11% ± 1.50 | 4 / 20 |
| Benefits Navigator | 65.41% ± 1.67 | 3 / 16 |