sabi-noema-oQ6e-mtp
This model was quantized from sabicap/sabi-noema using oQ (oMLX v0.7.1rc1) mixed-precision quantization.
The MTP head was grafted from Jundot/Qwen3.8-27B-oQ6e-mtp.
The effective average is 6.826 bits/weight for the complete checkpoint.
Quantization and Bit Distribution
| Storage format | Logical weights | Share of total | Tensor storage | Effective bits/weight |
|---|---|---|---|---|
| Affine 6-bit, group 64 | 25.184B | 90.650% | 19.057 GiB | 6.50 |
| Affine 8-bit, group 64 | 2.082B | 7.493% | 2.060 GiB | 8.50 |
| BF16 | 0.516B | 1.857% | 0.961 GiB | 16.00 |
| Total | 27.781B | 100% | 22.077 GiB | 6.826 |
Storage Breakdown
| Component | Logical weights | Storage |
|---|---|---|
| Language model | 26.896B | 22.377 GB / 20.840 GiB |
| Vision | 0.461B | 0.921 GB / 0.858 GiB |
| MTP head | 0.425B | 0.407 GB / 0.379 GiB |
| Total | 27.781B | 23.705 GB / 22.077 GiB |