Quantization formats
A quantization label names a mixture, not a uniform precision — so the effective bits per weight of a real file always exceeds the nominal rate its name implies. These are the measured averages across every file in the index.
| Format | Files measured | Effective bpw |
|---|---|---|
| UD-TQ1_0 | 50 | 2.067 |
| I1-IQ1_S | 957 | 2.068 |
| IQ1_S | 192 | 2.075 |
| UD-IQ1_S | 157 | 2.160 |
| IQ1_M | 255 | 2.170 |
| I1-IQ1_M | 956 | 2.209 |
| UD-IQ1_M | 178 | 2.311 |
| IQ2_XXS | 342 | 2.393 |
| I1-IQ2_XXS | 957 | 2.445 |
| UD-IQ2_XXS | 199 | 2.591 |
| IQ2_XS | 447 | 2.614 |
| I1-IQ2_XS | 957 | 2.647 |
| IQ2_S | 452 | 2.712 |
| I1-IQ2_S | 957 | 2.762 |
| TQ2_0 | 19 | 2.808 |
| UD-IQ2_M | 206 | 2.937 |
| I1-IQ2_M | 957 | 2.950 |
| IQ2_M | 745 | 2.988 |
| I1-Q2_K_S | 829 | 3.030 |
| Q2_K_S | 86 | 3.184 |
| I1-Q2_K | 1037 | 3.223 |
| UD-IQ3_S | 32 | 3.250 |
| TQ1_0 | 14 | 3.256 |
| I1-IQ3_XXS | 959 | 3.278 |
| UD-IQ3_XXS | 213 | 3.298 |
| Q2_K | 2331 | 3.337 |
| Q2_K_L | 764 | 3.379 |
| IQ3_XXS | 542 | 3.427 |
| Q2_K_M | 8 | 3.451 |
| I1-IQ3_XS | 960 | 3.518 |
| IQ3_XS | 788 | 3.562 |
| I1-Q3_K_S | 1036 | 3.680 |
| I1-IQ3_S | 1036 | 3.698 |
| Q3_K_S | 2241 | 3.744 |
| I1-IQ3_M | 1034 | 3.796 |
| IQ3_S | 235 | 3.867 |
| IQ3_M | 906 | 3.871 |
| UD-IQ4_XS | 36 | 3.989 |
| UD-IQ4_NL | 29 | 4.012 |
| I1-Q3_K_M | 1035 | 4.029 |
| Q3_K_M | 2461 | 4.127 |
| I1-Q3_K_L | 1034 | 4.321 |
| Q3_K_L | 2134 | 4.368 |
| UD-Q3_K_M | 37 | 4.412 |
| I1-IQ4_XS | 1034 | 4.413 |
| Q3_K | 84 | 4.493 |
| IQ4_XS | 1968 | 4.556 |
| I1-Q4_0 | 1033 | 4.625 |
| I1-Q4_K_S | 1034 | 4.672 |
| I1-IQ4_NL | 528 | 4.683 |
| UD-Q3_K_S | 10 | 4.737 |
| MXFP4 | 13 | 4.761 |
| IQ4_NL | 840 | 4.779 |
| Q4_K_S | 2360 | 4.795 |
| Q4_0 | 1376 | 4.913 |
| I1-Q4_K_M | 1033 | 4.925 |
| I1-Q4_1 | 897 | 5.051 |
| Q4_K_M | 3312 | 5.134 |
| Q4_K_L | 565 | 5.203 |
| UD-Q4_K_S | 28 | 5.285 |
| Q4_1 | 878 | 5.371 |
| I1-Q5_K_S | 1016 | 5.536 |
| Q5_K_S | 2319 | 5.656 |
| I1-Q5_K_M | 1018 | 5.691 |
| Q5_K_M | 2776 | 5.919 |
| UD-Q4_K_M | 32 | 5.947 |
| Q5_K_L | 527 | 5.964 |
| Q4_K | 164 | 6.044 |
| Q5_0 | 506 | 6.111 |
| UD-Q5_K_S | 28 | 6.255 |
| Q5_K | 108 | 6.374 |
| Q6_K_M | 10 | 6.415 |
| I1-Q6_K | 990 | 6.541 |
| UD-Q6_K | 28 | 6.740 |
| UD-Q2_K | 5 | 6.751 |
| Q6_K | 2999 | 6.808 |
| Q6_K_L | 526 | 6.811 |
| NVFP4 | 30 | 6.849 |
| Q5_1 | 268 | 6.942 |
| UD-Q5_K_M | 31 | 7.032 |
| Q8_0 | 3451 | 8.635 |
| Q2_0 | 7 | 8.699 |
| BF16 | 945 | 15.545 |
| F16 | 1161 | 15.704 |
| F32 | 158 | 31.257 |