Knowledge Evaluation on KMMLU, KMMLU Pro, GPQA Diamond, and Related Benchmarks
85.1AccuracyDeepSeek-V3.1
Evaluation Results
| Method | Links | |
|---|---|---|
| DeepSeek-V3.1Thinking Mode=true, Parameters=685B-A37B, Lang=English, Sub-Dataset=MMLU-Pro2026.01 | 85.1 | |
| A.X K1Thinking Mode=true, Parameters=519B-A33B, Lang=Korean, Sub-Dataset=CLIcK2026.01 | 84.9 | |
| GLM-4.6Thinking Mode=true, Parameters=357B-A32B, Lang=Korean, Sub-Dataset=CLIcK2026.01 | 84.9 | |
| DeepSeek-V3.1Thinking Mode=true, Parameters=685B-A37B, Lang=Korean, Sub-Dataset=CLIcK2026.01 | 84.5 | |
| GLM-4.6Thinking Mode=true, Parameters=357B-A32B, Lang=English, Sub-Dataset=MMLU-Pro2026.01 | 82.9 | |
| A.X K1Thinking Mode=true, Parameters=519B-A33B, Lang=English, Sub-Dataset=MMLU-Pro2026.01 | 81.5 | |
| A.X K1Thinking Mode=true, Parameters=519B-A33B, Lang=Korean, Sub-Dataset=KMMLU2026.01 | 80.2 | |
| GLM-4.6Thinking Mode=true, Parameters=357B-A32B, Lang=Korean, Sub-Dataset=KMMLU2026.01 | 79.9 | |
| GLM-4.6Thinking Mode=true, Parameters=357B-A32B, Lang=Korean, Sub-Dataset=KMMLU-Redux2026.01 | 78.2 | |
| GLM-4.6Thinking Mode=true, Parameters=357B-A32B, Lang=English, Sub-Dataset=GPQA Diamond2026.01 | 78 | |
| A.X K1Thinking Mode=true, Parameters=519B-A33B, Lang=Korean, Sub-Dataset=KMMLU-Redux2026.01 | 77.9 | |
| DeepSeek-V3.1Thinking Mode=true, Parameters=685B-A37B, Lang=English, Sub-Dataset=GPQA Diamond2026.01 | 77.9 | |
| DeepSeek-V3.1Thinking Mode=true, Parameters=685B-A37B, Lang=Korean, Sub-Dataset=KMMLU2026.01 | 76.5 | |
| DeepSeek-V3.1Thinking Mode=true, Parameters=685B-A37B, Lang=Korean, Sub-Dataset=KMMLU-Redux2026.01 | 75.9 | |
| A.X K1Thinking Mode=true, Parameters=519B-A33B, Lang=English, Sub-Dataset=GPQA Diamond2026.01 | 74 | |
| GLM-4.6Thinking Mode=true, Parameters=357B-A32B, Lang=Korean, Sub-Dataset=KMMLU-Pro2026.01 | 71.8 | |
| DeepSeek-V3.1Thinking Mode=true, Parameters=685B-A37B, Lang=Korean, Sub-Dataset=KMMLU-Pro2026.01 | 71.4 | |
| A.X K1Thinking Mode=true, Parameters=519B-A33B, Lang=Korean, Sub-Dataset=KMMLU-Pro2026.01 | 68.1 | |
| DeepSeek-V3.1Thinking Mode=true, Parameters=685B-A37B, Lang=Korean, Sub-Dataset=KoBALT2026.01 | 59.7 | |
| GLM-4.6Thinking Mode=true, Parameters=357B-A32B, Lang=Korean, Sub-Dataset=KoBALT2026.01 | 59.2 | |
| A.X K1Thinking Mode=true, Parameters=519B-A33B, Lang=Korean, Sub-Dataset=KoBALT2026.01 | 48.8 |