Quantization publisher

NexaAI

NexaAI publishes 44 quantizations across 3 models in our index, averaging 6.179 effective bits per weight. Their files differ in size from other publishers' builds of the same nominal quantization on 22 of the pairs we can compare — the same label does not mean the same file.

From the file· summed file bytes
Repositories
3
Quantizations
44
Models covered
3
Avg effective bpw
6.179
across their files

Same model, same quant label, different bytes

largest disagreements first
ModelQuantNexaAIvsTheirsDifference
gemma-2bQ2_K1.08 GiBbrittlewis120.84 GiB+28.7%
gemma-2bQ3_K_M1.29 GiBbrittlewis121.10 GiB+17.4%
gemma-2bQ3_K_L1.36 GiBbrittlewis121.17 GiB+16.2%
gemma-2bQ3_K_S1.20 GiBbrittlewis121.01 GiB+18.9%
gemma-2bQ4_K_S1.45 GiBbrittlewis121.33 GiB+9.5%
gemma-2bQ4_K_M1.52 GiBbrittlewis121.39 GiB+9.0%
gemma-2bQ5_K_S1.68 GiBbrittlewis121.61 GiB+4.0%
gemma-2bQ5_K_M1.71 GiBbrittlewis121.65 GiB+3.9%
DeepSeek-OCRQ8_02.90 GiBsabafallah2.91 GiB-0.3%
DeepSeek-OCRBF165.47 GiBsabafallah5.47 GiB-0.1%
gemma-2bQ8_02.49 GiBbrittlewis122.49 GiB-0.0%
gemma-2bQ6_K1.92 GiBbrittlewis121.92 GiB-0.0%
gemma-2bQ8_02.49 GiBMaziyarPanahi2.49 GiB0.0%
gemma-2bQ5_K_M1.71 GiBMaziyarPanahi1.71 GiB0.0%
gemma-2bQ3_K_L1.36 GiBMaziyarPanahi1.36 GiB0.0%
gemma-2bQ5_K_S1.68 GiBMaziyarPanahi1.68 GiB0.0%
gemma-2bQ6_K1.92 GiBMaziyarPanahi1.92 GiB0.0%
gemma-2bQ4_K_M1.52 GiBMaziyarPanahi1.52 GiB0.0%
gemma-2bQ2_K1.08 GiBMaziyarPanahi1.08 GiB0.0%
gemma-2bQ3_K_S1.20 GiBMaziyarPanahi1.20 GiB0.0%
gemma-2bQ4_K_S1.45 GiBMaziyarPanahi1.45 GiB0.0%
gemma-2bQ3_K_M1.29 GiBMaziyarPanahi1.29 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
gemma-2b181.08 GiB
DeepSeek-OCR81.54 GiB
octo-net181.32 GiB