Ling-3.0-tiny-oQ3

This model was quantized using oQ (oMLX v0.7.2) mixed-precision quantization.

The effective average is 3.700 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	7.565B	95.835%	3.082 GiB	3.50
Affine 4-bit, group 64	0.000B	0.002%	0.000 GiB	4.50
Affine 5-bit, group 64	0.013B	0.169%	0.009 GiB	5.50
Affine 6-bit, group 64	0.013B	0.171%	0.010 GiB	6.50
Affine 8-bit, group 64	0.301B	3.815%	0.298 GiB	8.50
BF16	0.001B	0.007%	0.001 GiB	16.00
FP32	0.000B	0.000%	0.000 GiB	32.00
Total	7.893B	100%	3.400 GiB	3.700

Storage Breakdown

Component	Logical weights	Storage
Language model	7.893B	3.651 GB / 3.400 GiB
Total	7.893B	3.651 GB / 3.400 GiB