Qwen2.5 VL 32B Instruct GGUF

🚀 Qwen2.5-VL-32B-Instruct GGUF Models

This project offers Qwen2.5-VL-32B-Instruct GGUF models, which are well - suited for image - text - to - text tasks. These models can handle multimodal data effectively, providing solutions for various applications.

🚀 Quick Start

How to Use Qwen 2.5 VL Instruct with llama.cpp (latest as of 10th May 2025)

1. Download the Qwen 2.5 VL gguf file:
   • Visit https://huggingface.co/Mungert/Qwen2.5-VL-32B-Instruct-GGUF/tree/main.
   • Choose a gguf file without the mmproj in the name. For example: https://huggingface.co/Mungert/Mungert/Qwen2.5-VL-32B-Instruct-GGUF/resolve/main/Qwen2.5-VL-32B-Instruct-q8_0.gguf.
   • Copy this file to your chosen folder.
2. Download the Qwen 2.5 VL mmproj file
   • Go to https://huggingface.co/Mungert/Qwen2.5-VL-32B-Instruct-GGUF/tree/main.
   • Select a file with mmproj in the name. For instance: https://huggingface.co/Mungert/Qwen2.5-VL-32B-Instruct-GGUF/resolve/main/Qwen2.5-VL-32B-Instruct-mmproj-f16.gguf.
   • Copy this file to your chosen folder.
3. Copy images to the same folder as the gguf files or alter paths appropriately.
   • In the example below, the gguf files, images, and llama - mtmd - cli are in the same folder.
   • Example image: https://huggingface.co/Mungert/Qwen2.5-VL-32B-Instruct-GGUF/resolve/main/car-1.jpg. Copy this file to your chosen folder.
4. Run the CLI Tool:
   • From your chosen folder, execute the following command:

llama-mtmd-cli -m Qwen2.5-VL-32B-Instruct-q8_0.gguf --mmproj Qwen2.5-VL-32B-Instruct-mmproj-f16.gguf  -p "Describe this image." --image ./car-1.jpg

✨ Features

Ultra - Low - Bit Quantization with IQ - DynamicGate (1 - 2 bit)

Our latest quantization method introduces precision - adaptive quantization for ultra - low - bit models (1 - 2 bit), with benchmark - proven improvements on Llama - 3 - 8B. This approach uses layer - specific strategies to preserve accuracy while maintaining extreme memory efficiency.

Benchmark Context

All tests were conducted on Llama - 3 - 8B - Instruct using:

• Standard perplexity evaluation pipeline
• 2048 - token context window
• The same prompt set across all quantizations

Method

• Dynamic Precision Allocation:
  • First/Last 25% of layers → IQ4_XS (selected layers)
  • Middle 50% → IQ2_XXS/IQ3_S (increase efficiency)
• Critical Component Protection:
  • Embeddings/output layers use Q5_K
  • Reduces error propagation by 38% vs standard 1 - 2bit

Quantization Performance Comparison (Llama - 3 - 8B)

Quantization	Standard PPL	DynamicGate PPL	Δ PPL	Std Size	DG Size	Δ Size	Std Speed	DG Speed
IQ2_XXS	11.30	9.84	-12.9%	2.5G	2.6G	+0.1G	234s	246s
IQ2_XS	11.72	11.63	-0.8%	2.7G	2.8G	+0.1G	242s	246s
IQ2_S	14.31	9.02	-36.9%	2.7G	2.9G	+0.2G	238s	244s
IQ1_M	27.46	15.41	-43.9%	2.2G	2.5G	+0.3G	206s	212s
IQ1_S	53.07	32.00	-39.7%	2.1G	2.4G	+0.3G	184s	209s

Key:

• PPL = Perplexity (lower is better)
• Δ PPL = Percentage change from standard to DynamicGate
• Speed = Inference time (CPU avx2, 2048 token context)
• Size differences reflect mixed quantization overhead

Key Improvements:

• 🔥 IQ1_M shows a massive 43.9% perplexity reduction (27.46 → 15.41)
• 🚀 IQ2_S cuts perplexity by 36.9% while adding only 0.2GB
• ⚡ IQ1_S maintains 39.7% better accuracy despite 1 - bit quantization

Tradeoffs:

• All variants have modest size increases (0.1 - 0.3GB)
• Inference speeds remain comparable (<5% difference)

When to Use These Models

📌 Fitting models into GPU VRAM ✔ Memory - constrained deployments ✔ Cpu and Edge Devices where 1 - 2bit errors can be tolerated ✔ Research into ultra - low - bit quantization

Choosing the Right Model Format

Selecting the correct model format depends on your hardware capabilities and memory constraints.

BF16 (Brain Float 16) – Use if BF16 acceleration is available

• A 16 - bit floating - point format designed for faster computation while retaining good precision.
• Provides similar dynamic range as FP32 but with lower memory usage.
• Recommended if your hardware supports BF16 acceleration (check your device’s specs).
• Ideal for high - performance inference with reduced memory footprint compared to FP32.

📌 Use BF16 if: ✔ Your hardware has native BF16 support (e.g., newer GPUs, TPUs). ✔ You want higher precision while saving memory. ✔ You plan to requantize the model into another format.

📌 Avoid BF16 if: ❌ Your hardware does not support BF16 (it may fall back to FP32 and run slower). ❌ You need compatibility with older devices that lack BF16 optimization.

F16 (Float 16) – More widely supported than BF16

• A 16 - bit floating - point high precision but with less of range of values than BF16.
• Works on most devices with FP16 acceleration support (including many GPUs and some CPUs).
• Slightly lower numerical precision than BF16 but generally sufficient for inference.

📌 Use F16 if: ✔ Your hardware supports FP16 but not BF16. ✔ You need a balance between speed, memory usage, and accuracy. ✔ You are running on a GPU or another device optimized for FP16 computations.

📌 Avoid F16 if: ❌ Your device lacks native FP16 support (it may run slower than expected). ❌ You have memory limitations.

Quantized Models (Q4_K, Q6_K, Q8, etc.) – For CPU & Low - VRAM Inference

Quantization reduces model size and memory usage while maintaining as much accuracy as possible.

• Lower - bit models (Q4_K) → Best for minimal memory usage, may have lower precision.
• Higher - bit models (Q6_K, Q8_0) → Better accuracy, requires more memory.

📌 Use Quantized Models if: ✔ You are running inference on a CPU and need an optimized model. ✔ Your device has low VRAM and cannot load full - precision models. ✔ You want to reduce memory footprint while keeping reasonable accuracy.

📌 Avoid Quantized Models if: ❌ You need maximum accuracy (full - precision models are better for this). ❌ Your hardware has enough VRAM for higher - precision formats (BF16/F16).

Very Low - Bit Quantization (IQ3_XS, IQ3_S, IQ3_M, Q4_K, Q4_0)

These models are optimized for extreme memory efficiency, making them ideal for low - power devices or large - scale deployments where memory is a critical constraint.

• IQ3_XS: Ultra - low - bit quantization (3 - bit) with extreme memory efficiency.
  • Use case: Best for ultra - low - memory devices where even Q4_K is too large.
  • Trade - off: Lower accuracy compared to higher - bit quantizations.
• IQ3_S: Small block size for maximum memory efficiency.
  • Use case: Best for low - memory devices where IQ3_XS is too aggressive.
• IQ3_M: Medium block size for better accuracy than IQ3_S.
  • Use case: Suitable for low - memory devices where IQ3_S is too limiting.
• Q4_K: 4 - bit quantization with block - wise optimization for better accuracy.
  • Use case: Best for low - memory devices where Q6_K is too large.
• Q4_0: Pure 4 - bit quantization, optimized for ARM devices.
  • Use case: Best for ARM - based devices or low - memory environments.

Summary Table: Model Format Selection

Model Format	Precision	Memory Usage	Device Requirements	Best Use Case
BF16	Highest	High	BF16 - supported GPU/CPUs	High - speed inference with reduced memory
F16	High	High	FP16 - supported devices	GPU inference when BF16 isn’t available
Q4_K	Medium Low	Low	CPU or Low - VRAM devices	Best for memory - constrained environments
Q6_K	Medium	Moderate	CPU with more memory	Better accuracy while still being quantized
Q8_0	High	Moderate	CPU or GPU with enough VRAM	Best accuracy among quantized models
IQ3_XS	Very Low	Very Low	Ultra - low - memory devices	Extreme memory efficiency and low accuracy
Q4_0	Low	Low	ARM or low - memory devices	llama.cpp can optimize for ARM devices

Included Files & Details

Qwen2.5-VL-32B-Instruct-bf16.gguf

• Model weights preserved in BF16.
• Use this if you want to requantize the model into a different format.
• Best if your device supports BF16 acceleration.

Qwen2.5-VL-32B-Instruct-f16.gguf

• Model weights stored in F16.
• Use if your device supports FP16, especially if BF16 is not available.

Qwen2.5-VL-32B-Instruct-bf16-q8_0.gguf

• Output & embeddings remain in BF16.
• All other layers quantized to Q8_0.
• Use if your device supports BF16 and you want a quantized version.

Qwen2.5-VL-32B-Instruct-f16-q8_0.gguf

• Output & embeddings remain in F16.
• All other layers quantized to Q8_0.

Qwen2.5-VL-32B-Instruct-q4_k.gguf

• Output & embeddings quantized to Q8_0.
• All other layers quantized to Q4_K.
• Good for CPU inference with limited memory.

Qwen2.5-VL-32B-Instruct-q4_k_s.gguf

• Smallest Q4_K variant, using less memory at the cost of accuracy.
• Best for very low - memory setups.

Qwen2.5-VL-32B-Instruct-q6_k.gguf

• Output & embeddings quantized to Q8_0.
• All other layers quantized to Q6_K.

Qwen2.5-VL-32B-Instruct-q8_0.gguf

• Fully Q8 quantized model for better accuracy.
• Requires more memory but offers higher precision.

Qwen2.5-VL-32B-Instruct-iq3_xs.gguf

• IQ3_XS quantization, optimized for extreme memory efficiency.
• Best for ultra - low - memory devices.

Qwen2.5-VL-32B-Instruct-iq3_m.gguf

• IQ3_M quantization, offering a medium block size for better accuracy.
• Suitable for low - memory devices.

Qwen2.5-VL-32B-Instruct-q4_0.gguf

• Pure Q4_0 quantization, optimized for ARM devices.
• Best for low - memory environments.
• Prefer IQ4_NL for better accuracy.

If you find these models useful

Please click like ❤. Also, I’d really appreciate it if you could test my Network Monitor Assistant at 👉 Network Monitor Assitant.

💬 Click the chat icon (bottom right of the main and dashboard pages). Choose an LLM; toggle between the LLM Types TurboLLM -> FreeLLM -> TestLLM.

What I'm Testing

I'm experimenting with function calling against my network monitoring service. Using small open - source models. I am into the question "How small can it go and still function".

🟡 TestLLM – Runs the current testing model using llama.cpp on 6 threads of a Cpu VM (Should take about 15s to load. Inference speed is quite slow and it only processes one user prompt at a time—still working on scaling!). If you're curious, I'd be happy to share how it works!

The other Available AI Assistants

🟢 TurboLLM – Uses gpt - 4o - mini Fast! Note: tokens are limited since OpenAI models are pricey, but you can Login or Download the Free Network Monitor agent to get more tokens, Alternatively use the TestLLM.

🔵 HugLLM – Runs open - source Hugging Face models Fast, Runs small models (≈8B) hence lower quality, Get 2x more tokens (subject to Hugging Face API availability)

📚 Documentation

Latest Updates

In addition to the original formula, we have further enhanced Qwen2.5-VL-32B's mathematical and problem - solving abilities through reinforcement learning. This has also significantly improved the model's subjective user experience, with response styles adjusted to better align with human preferences. Particularly for objective queries such as mathematics, logical reasoning, and knowledge - based Q&A, the level of detail in responses and the clarity of formatting have been noticeably enhanced.

Introduction

In the past five months since Qwen2 - VL’s release, numerous developers have built new models on the Qwen2 - VL vision - language models, providing us with valuable feedback. During this period, we focused on building more useful vision - language models. Today, we are excited to introduce the latest addition to the Qwen family: Qwen2.5 - VL.

Key Enhancements

• Understand things visually: Qwen2.5 - VL is not only proficient in recognizing common objects such as flowers, birds, fish, and insects, but it is highly capable of analyzing texts, charts, icons, graphics, and layouts within images.
• Being agentic: Qwen2.5 - VL directly plays as a visual agent that can reason and dynamically direct tools, which is capable of computer use and phone use.
• Understanding long videos and capturing events: Qwen2.5 - VL can comprehend videos of over 1 hour, and this time it has a new ability of capturing event by pinpointing the relevant video segments.
• Capable of visual localization in different formats: Qwen2.5 - VL can accurately localize objects in an image by generating bounding boxes or points, and it can provide stable JSON outputs for coordinates and attributes.
• Generating structured outputs: for data like scans of invoices, forms, tables, etc. Qwen2.5 - VL supports structured outputs of their contents, benefiting usages in finance, commerce, etc.

Model Architecture Updates

• Dynamic Resolution and Frame Rate Training for Video Understanding: We extend dynamic resolution to the temporal dimension by adopting dynamic FPS sampling, enabling the model to comprehend videos at various sampling rates. Accordingly, we update mRoPE in the time dimension with IDs and absolute time alignment, enabling the model to learn temporal sequence and speed, and ultimately acquire the ability to pinpoint specific moments.

* **Streamlined and Efficient Vision Encoder** We enhance both training and inference speeds by strategically implementing window attention into the ViT. The ViT architecture is further optimized with SwiGLU and RMSNorm, aligning it with the structure of the Qwen2.5 LLM.

We have three models with 3, 7, and 72 billion parameters. This repo contains the instruction - tuned 32B Qwen2.5 - VL model. For more information, visit our [Blog](https://qwenlm.github.io/blog/qwen2.5 - vl/) and [GitHub](https://github.com/QwenLM/Qwen2.5 - VL).

Evaluation

Vision

Dataset	Qwen2.5-VL-72B (🤗🤖)	Qwen2-VL-72B (🤗🤖)	Qwen2.5-VL-32B (🤗🤖)
MMMU	70.2	64.5	70
MMMU Pro	51.1	46.2	49.5
MMStar	70.8	68.3	69.5
MathVista	74.8	70.5	74.7
MathVision	38.1	25.9	40.0
OCRBenchV2	61.5/63.7	47.8/46.1	57.2/59.1
CC-OCR	79.8	68.7	77.1
DocVQA	96.4	96.5	94.8
InfoVQA	87.3	84.5	83.4
LVBench	47.3	-	49.00
CharadesSTA	50.9	-	54.2
VideoMME	73.3/79.1	71.2/77.8	70.5/77.9
MMBench-Video	2.02	1.7	1.93
AITZ	83.2	-	83.1
Android Control	67.4/93.7	66.4/84.4	69.6/93.3
ScreenSpot	87.1	-	88.5
ScreenSpot Pro	43.6	-	39.4
AndroidWorld	35	-	22.0
OSWorld	8.83	-	5.92

Text

MODEL	MMLU	MMLU - PRO	MATH	GPQA - diamond	MBPP	Human Eval
Qwen2.5 - VL - 32B	78.4	68.8	82.2	46.0	84.0	91.5
Mistral - Small - 3.1 - 24B	80.6	66.8	69.3	46.0	74.7	88.4
Gemma3 - 27B - IT	76.9	67.5	89	42.4	74.4	87.8
GPT - 4o - Mini	82.0	61.7	70.2	39.4	84.8	87.2
Claude - 3.5 - Haiku	77.6	65.0	69.2	41.6	85.6	88.1

💻 Usage Examples

Requirements

The code of Qwen2.5 - VL has been in the latest Hugging face transformers and we advise you to build from source with command:

pip install git+https://github.com/huggingface/transformers accelerate

or you might encounter the following error:

KeyError: 'qwen2_5_vl'

We offer a toolkit to help you handle various types of visual input more conveniently, as if you were using an API. This includes base64, URLs, and interleaved images and videos. You can install it using the following command:

# It's highly recommanded to use `[decord]` feature for faster video loading.
pip install qwen-vl-utils[decord]==0.0.8

If you are not using Linux, you might not be able to install decord from PyPI. In that case, you can use pip install qwen-vl-utils which will fall back to using torchvision for video processing. However, you can still install decord from source to get decord used when loading video.

Using 🤗 Transformers to Chat

from transformers import Qwen2_5_VLForConditionalGeneration, AutoTokenizer, AutoProcessor
from qwen_vl_utils import process_vision_info

# default: Load the model on the available device(s)
model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
    "Qwen/Qwen2.5-VL-32B-Instruct", torch_dtype="auto", device_map="auto"
)

# We recommend enabling flash_attention_2 for better acceleration and memory saving, especially in multi-image and video scenarios.
# model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
#     "Qwen/Qwen2.5-VL-32B-Instruct",
#     torch_dtype=torch.bfloat16,
#     attn_implementation="flash_attention_2",
#     device_map="auto",
# )

# default processer
processor = AutoProcessor.from_pretrained("Qwen/Qwen2.5-VL-32B-Instruct")

# The default range for the number of visual tokens per image in the model is 4-16384.
# You can set min_pixels and max_pixels according to your needs, such as a token range of 256-1280, to balance performance and cost.
# min_pixels = 256*28*28
# max_pixels = 1280*28*28
# processor = AutoProcessor.from_pretrained("Qwen/Qwen2.5-VL-32B-Instruct", min_pixels=min_pixels, max_pixels=max_pixels)

messages = [
    {
        "role": "user",
        "content": [
            {"name": "image0", "parameters": {"image": "your_image_path"}},
            "Describe this image."
        ]
    }
]

inputs = processor(text=messages, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=1024)
response = processor.decode(outputs[0], skip_special_tokens=True)
print(response)

📄 License

This project is licensed under the Apache - 2.0 license.