Qwen3.6-35B-A3B-oQ8e-mtp
This model was quantized using oQ (oMLX v0.7.1) mixed-precision quantization with Improved oQe Quantization (jundot/omlx#4385).
A mixed-precision MLX checkpoint of Qwen/Qwen3.6-35B-A3B, a mixture-of-experts vision-language model with 35B total and 3B activated parameters and a hybrid layout of Gated DeltaNet and Gated Attention layers. It includes the language model, the vision encoder and the multi-token prediction (MTP) head.
The effective average is 8.590 bits/weight for the complete checkpoint.
Quantization and Bit Distribution
oQ8e allocates bits by measured layer sensitivity, then rounds each group with an importance matrix (128 samples x 512 tokens) and a weighted least-squares refit of its scale and bias.
| Storage format | Logical weights | Share of total | Tensor storage | Effective bits/weight |
|---|---|---|---|---|
| Affine 8-bit, group 128 | 1.340B | 3.727% | 1.287 GiB | 8.25 |
| Affine 8-bit, group 64 | 34.134B | 94.945% | 33.777 GiB | 8.50 |
| BF16 | 0.478B | 1.329% | 0.890 GiB | 16.00 |
| Total | 35.952B | 100% | 35.953 GiB | 8.590 |
Storage Breakdown
| Component | Logical weights | Storage |
|---|---|---|
| Language backbone | 34.661B | 36.806 GB / 34.279 GiB |
| Vision encoder | 0.447B | 0.893 GB / 0.832 GiB |
| MTP head | 0.845B | 0.905 GB / 0.843 GiB |
| Total | 35.952B | 38.605 GB / 35.953 GiB |
License
Apache-2.0, following the source model. See LICENSE.