Ling-3.0-tiny-oQ2

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

The effective average is 3.000 bits/weight for the complete checkpoint.

Quantization and Bit Distribution

Storage format	Logical weights	Share of total	Tensor storage	Effective bits/weight
Affine 2-bit, group 64	7.230B	91.592%	2.104 GiB	2.50
Affine 3-bit, group 64	0.003B	0.042%	0.001 GiB	3.50
Affine 5-bit, group 64	0.007B	0.090%	0.005 GiB	5.50
Affine 6-bit, group 64	0.000B	0.000%	0.000 GiB	6.50
Affine 8-bit, group 64	0.653B	8.268%	0.646 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%	2.757 GiB	3.000

Storage Breakdown

Component	Logical weights	Storage
Language model	7.893B	2.960 GB / 2.757 GiB
Total	7.893B	2.960 GB / 2.757 GiB