mlx-community
/

Ministral-3-3B-Base-2512-6bit

Image-Text-to-Text

vision-language

Model card Files Files and versions

mabaeyens commited on 3 days ago

Commit

8eb94af

·

verified ·

1 Parent(s): 2d7efcd

Upload README.md with huggingface_hub

Files changed (1) hide show

README.md +1 -1

README.md CHANGED Viewed

@@ -67,7 +67,7 @@ description, capabilities, and license terms.
 - Source: [`mistralai/Ministral-3-3B-Base-2512`](https://huggingface.co/mistralai/Ministral-3-3B-Base-2512) (BF16)
 - Language model layers: **6-bit** affine quantization, group_size=64
 - Vision tower + multimodal projector: kept at full precision (not quantized)
-- Blended average: **24.790 bits per weight** across all parameters
 ## Ministral 3 family

 - Source: [`mistralai/Ministral-3-3B-Base-2512`](https://huggingface.co/mistralai/Ministral-3-3B-Base-2512) (BF16)
 - Language model layers: **6-bit** affine quantization, group_size=64
 - Vision tower + multimodal projector: kept at full precision (not quantized)
+- Blended average: **7.537 bits per weight** across all parameters
 ## Ministral 3 family