Reasoning on GPQA, MATH500, ARC-Challenge, MMLU
66GPQA Accuracygpt-oss-120b
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| gpt-oss-120bsize=120b2025.10 | 66 | 82 | 94 | 86 | 82 | |
| gemini-2.5-pro2025.10 | 42 | 36 | 94 | 80 | 63 | |
| deepseek-r12025.10 | 38 | 64 | 46 | 52 | 50 | |
| gpt-oss-20bsize=20b2025.10 | 32 | 18 | 62 | 54 | 42 | |
| gpt-oss-safeguardvariant=safeguard2025.10 | 20 | 42 | 69 | 66 | 49 |