LLM Evaluation Suite MMLU, GSM8k, HellaSwag, WinoGrande
64.43MMLU ScoreFP16
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| FP16Format=-2025.09 | 64.43 | 78.01 | 73.42 | 70.09 | 71.49 | — | |
| GPTQFormat=FP82025.09 | 64.1 | 77.8 | 73 | 69.9 | 71.2 | 99.54 | |
| RTNFormat=INT82025.09 | 64 | 77.7 | 73.3 | 69.6 | 71.15 | 99.47 | |
| GPTQFormat=INT82025.09 | 64 | 77.9 | 73.4 | 69.9 | 71.3 | 99.67 | |
| RTNFormat=FP82025.09 | 63.4 | 77.7 | 73 | 69.7 | 70.95 | 99.19 | |
| QATFormat=NVFP2025.09 | 62.06 | 75.06 | 71.27 | 67.96 | 69.09 | 96.64 | |
| QAT+Had16Format=NVFP2025.09 | 62.03 | 72.93 | 70.95 | 66.46 | 68.09 | 95.25 | |
| GPTQFormat=NVFP2025.09 | 61.76 | 70.36 | 71.07 | 69.93 | 68.28 | 95.51 | |
| MR-GPTQFormat=NVINT42025.09 | 60.8 | 72.25 | 71.74 | 70.48 | 68.82 | 96.27 | |
| MicroQAT+Had16Format=NVFP2025.09 | 60.66 | 69.98 | 70.55 | 67.01 | 67.05 | 93.79 | |
| RTNFormat=NVFP2025.09 | 60.62 | 70.43 | 70.99 | 68.03 | 67.52 | 94.45 | |
| GPTQFormat=NVINT42025.09 | 60.5 | 71.42 | 27.01 | 51.07 | 52.5 | 73.44 | |
| GPTQ+Had16Format=NVFP2025.09 | 60.26 | 68.76 | 71.05 | 67.8 | 66.97 | 93.68 | |
| RTNFormat=NVINT42025.09 | 60.22 | 71.65 | 70.92 | 66.77 | 67.39 | 94.27 | |
| GPTQ+Had128Format=NVFP2025.09 | 60.19 | 70.89 | 70.97 | 68.19 | 67.56 | 94.51 | |
| RTN+Had16Format=NVFP2025.09 | 59.91 | 64.82 | 69.77 | 65.59 | 65.02 | 90.96 | |
| QAT+Had32Format=MXFP2025.09 | 59.83 | 72.48 | 70.27 | 66.54 | 67.28 | 94.11 | |
| GPTQ+Had32Format=MXFP2025.09 | 59.79 | 68.92 | 69.5 | 66.85 | 66.27 | 92.69 | |
| GPTQ+Had128Format=MXFP2025.09 | 59.56 | 67.78 | 70.08 | 68.03 | 66.36 | 92.83 | |
| MicroQAT+Had32Format=MXFP2025.09 | 59.49 | 65.66 | 69.05 | 67.32 | 65.38 | 91.46 | |
| MR-GPTQFormat=MXINT42025.09 | 57.73 | 67.48 | 68.92 | 66.69 | 65.21 | 91.21 | |
| GPTQFormat=MXFP2025.09 | 57.68 | 62.32 | 63.87 | 64.88 | 62.19 | 86.99 | |
| RTNFormat=MXFP2025.09 | 56.81 | 60.8 | 67.3 | 64.56 | 62.37 | 87.24 | |
| RTN + HTFormat=NVINT42025.09 | 56.75 | 71.95 | 69.24 | 66.61 | 66.14 | 92.52 | |
| QATFormat=MXFP2025.09 | 56.17 | 64.9 | 69.51 | 67.17 | 64.44 | 90.14 | |
| RTN+Had128Format=MXFP2025.09 | 55.95 | 60.8 | 67.57 | 64.88 | 62.3 | 87.15 | |
| RTN+Had32Format=MXFP2025.09 | 55.58 | 57.77 | 68.56 | 64.33 | 61.56 | 86.11 | |
| RTN+Had128Format=NVFP2025.09 | 54.34 | 67.48 | 69.69 | 66.93 | 64.61 | 90.38 | |
| GPTQFormat=MXINT42025.09 | 53.48 | 60.73 | 51.44 | 58.8 | 56.11 | 78.49 | |
| RTN + HTFormat=MXINT42025.09 | 50.55 | 51.33 | 64.61 | 59.83 | 56.58 | 79.15 | |
| RTNFormat=MXINT42025.09 | 46.03 | 47.54 | 64.21 | 61.56 | 54.84 | 76.71 |