Anthropic's most powerful model at launch, with expert-level performance across math, coding, and reasoning.
| Benchmark | Description | Score | Category | Bar |
|---|---|---|---|---|
| MMLU | Massive Multitask Language Understanding | 86.8% | knowledge | |
| HumanEval | Code generation accuracy | 84.9% | coding | |
| MATH | Mathematical reasoning | 87.3% | math | |
| GSM8K | Grade school math | 87.3% | math | |
| GPQA | Graduate-level science Q&A | 58.5% | reasoning | |
| HellaSwag | Common sense reasoning | 86.7% | reasoning | |
| SWE-bench | Software engineering problems | 37.7% | coding | |
| ARC-C | AI2 Reasoning Challenge | 94% | reasoning |