Downloads · 30 days
1.5K
9% of all-time downloads
KuroTo4ka/Qwen3-VL-8B-Instruct-Unredacted-MAX-Quants-GGUF
Qwen3-VL-8B-Instruct-Unredacted-MAX-Quants-GGUF is a image-to-text model from KuroTo4ka. Use it when you need a caption or text from an image. It is set up for gguf. The card lists the license as apache-2.0.
Downloads · 30 days
1.5K
9% of all-time downloads
All-time downloads
16.6K
Public
Repo size
70.4 GB
Likes
12
Public
Click a slice to open those files.
.gguf70.4 GB · 100%
From the Hugging Face model README
This repository contains high-quality GGUF quantizations for the prithivMLmods/Qwen3-VL-8B-Instruct-Unredacted-MAX model.
mmproj) for different hardware needs.llama.cpp and other GGUF-supported backends.| Filename | Quant Method | Description |
|---|---|---|
Q4_K_M.gguf | Q4_K_M | Recommended. Best balance of speed and intelligence. |
Q8_0.gguf | Q8_0 | High quality, nearly identical to original weights. |
Q6_K.gguf | Q6_K | Very high quality, slightly slower than Q4. |
Q5_K_M.gguf | Q5_K_M | Good balance between Q4 and Q6. |
Q3_K_M.gguf | Q3_K_M | Low size, moderate quality loss. |
Q2_K.gguf | Q2_K | Smallest possible size, significant quality loss. |
F16.gguf | F16 | Baseline reference quality. |
Required for image recognition tasks.
| Filename | Type | Description |
|---|---|---|
mmproj-f32.gguf | F32 | Absolute maximum precision (2.3GB). |
mmproj-f16.gguf | F16 | Industry standard for high-quality vision. |
mmproj-bf16.gguf | BF16 | Optimized for modern NVIDIA GPUs (Ampere+). |
mmproj-q8_0.gguf | Q8_0 | Best for saving VRAM without losing recognition detail. |
To use vision capabilities in llama.cpp, use the following command:
./llama-cli -m Qwen3-VL-8B-Instruct-Unredacted-MAX.Q4_K_M.gguf \
--mmproj Qwen3-VL-8B-Instruct-Unredacted-MAX.mmproj-f16.gguf \
--image path/to/your/image.jpg \
-p "Describe this image"