Qwen2.5-72B-Instruct-GGUF / perplexity.md
ThomasBaruzier's picture
Upload perplexity.md
9ce8e86 verified

Qwen2.5-72B-Instruct Quant Size (MB) PPL Size (%) Accuracy (%) PPL error rate IQ1_S 21640 7.6552 0.10700 IQ1_M 22641 7.2982 0.10210 IQ2_XXS 24310 6.3958 0.08698 IQ2_XS 25805 6.0909 0.08248 IQ2_S 26645 6.0318 0.08180 IQ2_M 27980 5.7589 0.07721 Q2_K_S 28200 5.9731 0.08266 Q2_K 28431 5.9188 0.08204 IQ3_XXS 30370 5.5227 0.07426 IQ3_XS 31321 5.4357 0.07228 IQ3_S 32891 5.3782 0.07153 Q3_K_S 32891 5.4492 0.07429 IQ3_M 33859 5.3550 0.07069 Q3_K_M 35953 5.4069 0.07356 Q3_K_L 37676 5.4116 0.07371 IQ4_XS 37870 5.2776 0.07108 IQ4_NL 39402 5.2747 0.07099 Q4_0 39467 5.2998 0.07117 Q4_K_S 41857 5.2535 0.07066 Q4_1 43581 5.2801 0.07092 Q4_K_M 45220 5.2478 0.07054