Premium reasoning model for high-stakes tasks requiring the highest accuracy. Offers extended thinking with up to 100K tokens of reasoning.
| Benchmark | Description | Score | Category | Bar |
|---|---|---|---|---|
| MMLU | Massive Multitask Language Understanding | 87.3% | knowledge | |
| HumanEval | Code generation accuracy | 91.8% | coding | |
| MATH | Mathematical reasoning | 96.4% | math | |
| GSM8K | Grade school math | 96.7% | math | |
| GPQA | Graduate-level science Q&A | 77.6% | reasoning | |
| HellaSwag | Common sense reasoning | 88.6% | reasoning | |
| SWE-bench | Software engineering problems | 44% | coding | |
| ARC-C | AI2 Reasoning Challenge | 97% | reasoning |