Quantization publisher

Mindrend

Mindrend publishes 8 quantizations across 1 models in our index, averaging 5.168 effective bits per weight. Their files differ in size from other publishers' builds of the same nominal quantization on 8 of the pairs we can compare — the same label does not mean the same file.

From the file· summed file bytes
Repositories
1
Quantizations
8
Models covered
1
Avg effective bpw
5.168
across their files

Same model, same quant label, different bytes

largest disagreements first
ModelQuantMindrendvsTheirsDifference
dolphin-2.7-mixtral-8x7bQ2_K14.57 GiBsecond-state14.57 GiB0.0%
dolphin-2.7-mixtral-8x7bQ3_K_M18.96 GiBsecond-state18.96 GiB0.0%
dolphin-2.7-mixtral-8x7bQ4_024.63 GiBsecond-state24.63 GiB0.0%
dolphin-2.7-mixtral-8x7bQ4_K_M24.63 GiBsecond-state24.63 GiB0.0%
dolphin-2.7-mixtral-8x7bQ5_030.02 GiBsecond-state30.02 GiB0.0%
dolphin-2.7-mixtral-8x7bQ5_K_M30.02 GiBsecond-state30.02 GiB0.0%
dolphin-2.7-mixtral-8x7bQ6_K35.74 GiBsecond-state35.74 GiB0.0%
dolphin-2.7-mixtral-8x7bQ8_046.22 GiBsecond-state46.22 GiB0.0%

A quantization label describes a target, not a recipe. Publishers make different choices about which tensors to keep at higher precision, and some apply an importance matrix while others don't — so two files both honestly labelled the same thing can differ measurably in size and in quality.

Models they publish

ModelQuantizationsSmallest
dolphin-2.7-mixtral-8x7b814.57 GiB