---
base_model:
- Qwen/Qwen3.5-397B-A17B-FP8
tags:
- mxfp4
- rdna4
- amd
metrics:
- accuracy
- code_eval
pipeline_tag: image-text-to-text
library_name: transformers
---

this is really mxfp4 model tested on R9700 and got 33t/s. Best model for 256GB vram

Mex Context Size: 131k