djdeniro
/

Qwen3.5-397B-A17B-MXFP4

Image-Text-to-Text

8-bit precision

compressed-tensors

Model card Files Files and versions

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

Mex Context Size: 131k

Downloads last month: 262

Safetensors

Model size

215B params

Tensor type

BF16

·

U8

·

Model tree for djdeniro/Qwen3.5-397B-A17B-MXFP4

Base model

Qwen/Qwen3.5-397B-A17B

Quantized

Qwen/Qwen3.5-397B-A17B-FP8

Quantized

(5)

this model