Huihui-Qwen3.8-27B-abliterated-oQ4e
Updated 2026-11-10: Re-quantized with oMLX v0.7.4.dev1. Please re-download if you have an earlier copy.
This model was quantized using oQ (oMLX v0.7.4.dev1) mixed-precision quantization.
The effective average is 5.028 bits/weight for the complete checkpoint.
Quantization and Bit Distribution
| Storage format | Logical weights | Share of total | Tensor storage | Effective bits/weight |
|---|---|---|---|---|
| Affine 4-bit, group 64 | 19.063B | 69.683% | 9.987 GiB | 4.50 |
| Affine 5-bit, group 64 | 6.547B | 23.932% | 4.192 GiB | 5.50 |
| Affine 6-bit, group 64 | 1.282B | 4.687% | 0.970 GiB | 6.50 |
| Affine 8-bit, group 64 | 0.001B | 0.004% | 0.001 GiB | 8.50 |
| BF16 | 0.463B | 1.694% | 0.863 GiB | 16.00 |
| Total | 27.357B | 100% | 16.013 GiB | 5.028 |
Storage Breakdown
| Component | Logical weights | Storage |
|---|---|---|
| Language model | 26.896B | 16.272 GB / 15.155 GiB |
| Vision | 0.461B | 0.921 GB / 0.858 GiB |
| Total | 27.357B | 17.194 GB / 16.013 GiB |
- Downloads last month
- 73
Model size
27B params
Tensor type
U32
·
BF16 ·
Hardware compatibility
Log In to add your hardware
4-bit
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support