Qwen3.8-Flash-Next-oQ6e-mtp
[!NOTE] Updated 2026-10-29: Re-quantized with the improved oQe quantization in oMLX 0.7.3. Please re-download if you have an earlier copy.
This model was quantized using oQ (oMLX v0.7.3) mixed-precision quantization with improved oQe (jundot/omlx#4930).
A mixed-precision MLX checkpoint of Qwen/Qwen3.8-Flash-Next, an experimental preview of the architecture behind Qwen4: a mixture-of-experts vision-language model with 125B parameters and 6B activated, plus a 51B n-gram embedding and a 4B MTP module, that pairs Gated DeltaNet with Qwen Sparse Attention (QSA) and uses Gated Residual streams. It includes the language model, the n-gram embedding, the vision encoder and the multi-token prediction (MTP) head.
The effective average is 6.582 bits/weight for the language backbone and 6.724 bits/weight for the complete checkpoint.
Quantization and Bit Distribution
oQ6e allocates bits by measured layer sensitivity, then rounds each group with an importance matrix (1024 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 6-bit, group 64 | 123.313B | 68.507% | 93.311 GiB | 6.50 |
| Affine 8-bit, group 128 | 0.241B | 0.134% | 0.231 GiB | 8.25 |
| Affine 8-bit, group 64 | 4.731B | 2.628% | 4.681 GiB | 8.50 |
| I64 | 0.000B | 0.000% | 0.000 GiB | 64.00 |
| BF16 | 0.516B | 0.287% | 0.961 GiB | 16.00 |
| N-gram affine 6-bit, group 32 | 51.200B | 28.445% | 41.723 GiB | 7.00 |
| Total | 180.000B | 100% | 140.907 GiB | 6.724 |
Storage Breakdown
| Component | Logical weights | Storage |
|---|---|---|
| Language backbone | 125.744B | 103.458 GB / 96.352 GiB |
| N-gram embedding | 51.200B | 44.800 GB / 41.723 GiB |
| Vision encoder | 0.449B | 0.898 GB / 0.836 GiB |
| MTP head | 2.607B | 2.142 GB / 1.995 GiB |
| Total | 180.000B | 151.298 GB / 140.907 GiB |
License
Qwen Community License 1.0, following the source model. See LICENSE.