corpus Qa1m
6 models evaluated
|
| 1 | DeepSeek V4 Pro (Max) | DeepSeek | Open | 62.0% | |
| 2 | DeepSeek V4 Flash (Max) | DeepSeek | Open | 60.5% | |
| 3 | DeepSeek V4 Flash (High) | DeepSeek | Open | 59.3% | |
| 4 | DeepSeek V4 Pro (High) | DeepSeek | Open | 56.5% | |
| 5 | DeepSeek V4 Pro | DeepSeek | Open | 35.6% | |
| 6 | DeepSeek V4 Flash | DeepSeek | Open | 15.5% | |