clef-oQ4e
[!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 Cloudflare/clef, a 27B decision model post-trained from Qwen/Qwen3.8-27B that turns a state and a schema of typed questions into decisions, returning a probability for every allowed option of every question in a single forward pass. It includes the Qwen3.8-27B language backbone, the vision encoder and the joint schema head.
The joint schema head (joint_head.safetensors, joint_head_config.json) is copied unchanged from the source. oMLX serves this checkpoint through POST /v1/systemone.
The effective average is 4.840 bits/weight for the language backbone and 5.079 bits/weight for the complete checkpoint.
Quantization and Bit Distribution
oQ4e 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 4-bit, group 64 | 19.063B | 69.359% | 9.987 GiB | 4.50 |
| Affine 5-bit, group 64 | 6.547B | 23.821% | 4.192 GiB | 5.50 |
| Affine 6-bit, group 64 | 1.282B | 4.665% | 0.970 GiB | 6.50 |
| Affine 8-bit, group 64 | 0.001B | 0.004% | 0.001 GiB | 8.50 |
| BF16 | 0.591B | 2.152% | 1.102 GiB | 16.00 |
| Total | 27.485B | 100% | 16.251 GiB | 5.079 |
Storage Breakdown
| Component | Logical weights | Storage |
|---|---|---|
| Language backbone | 26.896B | 16.272 GB / 15.155 GiB |
| Vision encoder | 0.461B | 0.921 GB / 0.858 GiB |
| Joint schema head | 0.128B | 0.256 GB / 0.239 GiB |
| Total | 27.485B | 17.450 GB / 16.251 GiB |
License
Apache-2.0, following the source model. See LICENSE.