Claude Opus 4.5
Premium model combining maximum intelligence with practical performance. Best model in the world for coding, agents, and computer use. Most robustly aligned model with best prompt injection resistance of any frontier model. Features extended thinking, 200K context window, 64K max output, and a new effort parameter for controlling reasoning depth. Pricing: $5/$25 per million tokens (input/output).
Benchmark results
| Benchmark | Score | Tags | Source |
|---|---|---|---|
| ARC-AGI v2 | 37.6% | self-reported llm-stats | link → |
| GPQA | 87.0% | self-reported llm-stats | link → |
| MCP Atlas | 62.3% | self-reported llm-stats | link → |
| MMMLU | 90.8% | self-reported llm-stats | link → |
| MMMU (validation) | 80.7% | self-reported llm-stats | link → |
| OSWorld | 66.3% | self-reported llm-stats | link → |
| SWE-Bench Verified | 80.9% | self-reported llm-stats | link → |
| Tau2 Retail | 88.9% | self-reported llm-stats | link → |
| Tau2 Telecom | 98.2% | self-reported llm-stats | link → |
| Terminal-Bench 2.0 | 59.3% | self-reported llm-stats | link → |