MLX MXFP4 quantization of ArliAI/gpt-oss-20b-Derestricted.
reasoning_effort parameterMatches OpenAI's original MXFP4 quantization scheme:
| Component | Bits | Group Size | Format |
|---|---|---|---|
| MLP Experts | 4 | 32 | MXFP4 |
| Attention | - | - | Full precision (bfloat16) |
| Routers | - | - | Full precision (bfloat16) |
| Embeddings | - | - | Full precision (bfloat16) |
| LM Head | - | - | Full precision (bfloat16) |
mlx_lm.chat --model txgsync/gpt-oss-20b-Derestricted-mxfp4-mlx
For full Reasoning Effort support in LM Studio, install via:
lms get txgsync/gpt-oss-20b-derestricted
This downloads the model with the virtual model wrapper that enables the Reasoning Effort selector.
modules_to_not_convert scheme for optimal qualitymodel-00001-of-00003.safetensors 5.3 GBmodel-00002-of-00003.safetensors 5.2 GBmodel-00003-of-00003.safetensors 3.2 GB