|
Qwen2.5-32B-Instruct |
|
Quant Size (MB) PPL Size (%) Accuracy (%) PPL error rate |
|
IQ1_S 6938 12.2991 0.08384 |
|
IQ1_M 7565 10.2638 0.06990 |
|
IQ2_XS 9497 7.3601 0.04846 |
|
IQ2_S 9907 7.2397 0.04762 |
|
IQ2_M 10743 6.7268 0.04354 |
|
Q2_K_S 10956 6.9981 0.04644 |
|
Q2_K 11743 6.6603 0.04324 |
|
IQ3_XXS 12245 6.1570 0.03929 |
|
IQ3_XS 13071 6.0366 0.03833 |
|
Q3_K_S 13726 6.0878 0.03872 |
|
IQ3_S 13769 5.9886 0.03816 |
|
IQ3_M 14125 5.9942 0.03802 |
|
Q3_K_M 15197 5.8008 0.03677 |
|
Q3_K_L 16449 5.7812 0.03667 |
|
IQ4_XS 16874 5.6502 0.03586 |
|
IQ4_NL 17817 5.6408 0.03575 |
|
Q4_0 17845 5.6946 0.03599 |
|
Q4_K_S 17915 5.6367 0.03561 |
|
Q4_K_M 18932 5.6224 0.03554 |
|
IQ2_XXS 8611 8.0187 0.05388 |
|
Q4_1 19684 5.6586 0.03587 |
|
Q5_K_S 21590 5.5680 0.03515 |
|
Q5_0 21658 5.5880 0.03538 |
|
Q5_K_M 22185 5.5670 0.03515 |
|
Q5_1 23496 5.5734 0.03520 |
|
Q6_K 25641 5.5305 0.03483 |
|
Q8_0 33208 5.5221 0.03478 |
|
F16 62500 5.5191 0.03474 |
|
|