MiniMax-M3-oQ4e
This model was quantized using oQ (oMLX v0.7.3) mixed-precision quantization.
The effective average is 4.648 bits/weight for the model weights, or 4.669 bits/weight including the DSpark draft.
Quantization and Bit Distribution
| Storage format | Logical weights | Share of total | Tensor storage | Effective bits/weight |
|---|---|---|---|---|
| Affine 4-bit, group 64 | 413.122B | 96.741% | 216.422 GiB | 4.50 |
| Affine 8-bit, group 64 | 13.007B | 3.046% | 12.871 GiB | 8.50 |
| BF16 | 0.866B | 0.203% | 1.614 GiB | 16.00 |
| FP32 | 0.045B | 0.010% | 0.167 GiB | 32.00 |
| Total | 427.040B | 100% | 231.073 GiB | 4.648 |
Storage Breakdown
| Component | Logical weights | Storage |
|---|---|---|
| Language model | 426.175B | 246.382 GB / 229.461 GiB |
| Vision | 0.866B | 1.731 GB / 1.612 GiB |
| DSpark draft (MTP) | 2.011B | 2.302 GB / 2.144 GiB |
| Total | 429.051B | 250.415 GB / 233.217 GiB |
- Downloads last month
- -
Model size
427B params
Tensor type
U32
·
BF16 ·
F32 ·
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