embeddinggemma-2-oQ3

This model was quantized using oQ (oMLX v0.7.1) mixed-precision quantization.

The effective average is 11.625 bits/weight for the complete checkpoint.

Quantization and Bit Distribution

Storage format	Logical weights	Share of total	Tensor storage	Effective bits/weight
Affine 3-bit, group 64	0.219B	29.441%	0.089 GiB	3.50
Affine 5-bit, group 64	0.037B	5.001%	0.024 GiB	5.50
Affine 6-bit, group 64	0.004B	0.563%	0.003 GiB	6.50
Affine 8-bit, group 64	0.012B	1.550%	0.011 GiB	8.50
BF16	0.472B	63.445%	0.880 GiB	16.00
Total	0.744B	100%	1.007 GiB	11.625

Storage Breakdown

Component	Logical weights	Storage
Language model	0.272B	0.137 GB / 0.128 GiB
Vision	0.167B	0.335 GB / 0.312 GiB
Audio	0.305B	0.610 GB / 0.568 GiB
Total	0.744B	1.082 GB / 1.007 GiB