LLM Evaluation on Qwen3-1.7B Evaluation Suite (avg)
58.64Average PerformanceBF16
Evaluation Results
| Method | Links | |
|---|---|---|
| BF16Quantization Bit-width (W-A-KV)=W16-A16-KV162026.04 | 58.64 | |
| EDGERAZORQuantization Bit-width (W-A-KV)=W4-A16-KV162026.04 | 58.56 | |
| AutoRoundQuantization Bit-width (W-A-KV)=W4-A16-KV162026.04 | 58.31 | |
| AQLMQuantization Bit-width (W-A-KV)=W4-A16-KV162026.04 | 57.57 | |
| GPTAQQuantization Bit-width (W-A-KV)=W4-A16-KV162026.04 | 57.05 | |
| AWQQuantization Bit-width (W-A-KV)=W4-A16-KV162026.04 | 56.95 | |
| VPTQQuantization Bit-width (W-A-KV)=W4-A16-KV162026.04 | 56.52 | |
| GPTQQuantization Bit-width (W-A-KV)=W4-A16-KV162026.04 | 54.94 | |
| EDGERAZORQuantization Bit-width (W-A-KV)=W3-A16-KV162026.04 | 53.33 | |
| AutoRoundQuantization Bit-width (W-A-KV)=W3-A16-KV162026.04 | 51.48 | |
| AQLMQuantization Bit-width (W-A-KV)=W3-A16-KV162026.04 | 51.24 | |
| Q-PaletteQuantization Bit-width (W-A-KV)=W4-A16-KV162026.04 | 49.77 | |
| AWQQuantization Bit-width (W-A-KV)=W3-A16-KV162026.04 | 47.71 | |
| Q-PaletteQuantization Bit-width (W-A-KV)=W3-A16-KV162026.04 | 47.66 | |
| VPTQQuantization Bit-width (W-A-KV)=W3-A16-KV162026.04 | 47.44 | |
| EDGERAZORQuantization Bit-width (W-A-KV)=W2-A16-KV162026.04 | 47.14 | |
| Slim-LLM+Quantization Bit-width (W-A-KV)=W3-A16-KV162026.04 | 46.8 | |
| QTIPQuantization Bit-width (W-A-KV)=W2-A16-KV162026.04 | 45.85 | |
| OmniQuantQuantization Bit-width (W-A-KV)=W4-A16-KV162026.04 | 44.76 | |
| GPTAQQuantization Bit-width (W-A-KV)=W3-A16-KV162026.04 | 44.45 | |
| EDGERAZORQuantization Bit-width (W-A-KV)=W1.58-A16-KV162026.04 | 43.89 | |
| GPTQQuantization Bit-width (W-A-KV)=W3-A16-KV162026.04 | 43.14 | |
| AQLMQuantization Bit-width (W-A-KV)=W2-A16-KV162026.04 | 41.44 | |
| OmniQuantQuantization Bit-width (W-A-KV)=W3-A16-KV162026.04 | 40.65 | |
| AutoRoundQuantization Bit-width (W-A-KV)=W2-A16-KV162026.04 | 35.27 | |
| VPTQQuantization Bit-width (W-A-KV)=W2-A16-KV162026.04 | 33.13 | |
| Q-PaletteQuantization Bit-width (W-A-KV)=W2-A16-KV162026.04 | 33.06 | |
| QuIP#Quantization Bit-width (W-A-KV)=W3-A16-KV162026.04 | 32.88 | |
| QuIP#Quantization Bit-width (W-A-KV)=W4-A16-KV162026.04 | 32.67 | |
| OmniQuantQuantization Bit-width (W-A-KV)=W2-A16-KV162026.04 | 32.62 | |
| Slim-LLM+Quantization Bit-width (W-A-KV)=W2-A16-KV162026.04 | 32.26 | |
| QuIP#Quantization Bit-width (W-A-KV)=W2-A16-KV162026.04 | 31.33 | |
| Q-PaletteQuantization Bit-width (W-A-KV)=W1.58-A16-KV162026.04 | 30.92 | |
| AWQQuantization Bit-width (W-A-KV)=W2-A16-KV162026.04 | 30.85 | |
| ARB-LLMQuantization Bit-width (W-A-KV)=W1.58-A16-KV162026.04 | 30.61 | |
| GPTAQQuantization Bit-width (W-A-KV)=W2-A16-KV162026.04 | 29.98 | |
| GPTQQuantization Bit-width (W-A-KV)=W2-A16-KV162026.04 | 29.82 | |
| BiLLMQuantization Bit-width (W-A-KV)=W1.58-A16-KV162026.04 | 29.15 |