QW

Qwen2.5-VL-72B-Instruct

QwenGGUF by MungertUpdated 2025-09-24
Download on Hugging Face
Parameters73B
Context125K
Layers80
Architectureqwen2vl
Smallest quant24.2 GB
Quick answer

Qwen2.5-VL-72B-Instruct needs about 50.9 GB of memory at Q4_K_M (a 47.7 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 28 quants, from 24.2 GB (IQ1_M) to 145.4 GB (F16). Longer context needs more: at 32K tokens the KV cache alone adds about 10.7 GB.

Qwen2.5-VL-72B-Instruct: 72B multimodal instruction model

Qwen2.5-VL-72B-Instruct is a 72‑billion‑parameter vision‑language model that supports image‑to‑text and video understanding. It can analyze text, charts, icons, and layouts in images, generate structured JSON outputs, and localize objects with bounding boxes. The model is available in multiple GGUF quantizations.

Good at

  • Image and video analysis
  • Structured output generation
  • Dynamic resolution and frame‑rate video understanding

Limits

  • Requires large VRAM for higher‑bit quantizations
  • No vision support in lower‑bit IQ1_M
  • License is other, not open source

Best for

  • Multimodal chat assistants
  • Document and invoice parsing
  • Video event detection

Who it’s for. Developers building multimodal applications that need detailed image and video analysis. Researchers requiring structured output from visual data. Teams that can allocate sufficient GPU memory for large‑scale inference.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ1_M24.2 GB 27.4 GB
Too big Get file →
IQ2_XXS25.9 GB 29.1 GB
Too big Get file →
IQ2_XS27.5 GB 30.7 GB
Too big Get file →
IQ2_S28.3 GB 31.4 GB
Too big Get file →
IQ2_M29.7 GB 32.8 GB
Too big Get file →
Q2_K_S29.9 GB 33.0 GB
Too big Get file →
IQ3_XXS32.2 GB 35.4 GB
Too big Get file →
IQ3_XS33.0 GB 36.2 GB
Too big Get file →
IQ3_S34.6 GB 37.8 GB
Too big Get file →
Q3_K_S34.6 GB 37.8 GB
Too big Get file →
IQ3_M35.7 GB 38.8 GB
Too big Get file →
Q3_K_M37.9 GB 41.0 GB
Too big Get file →
IQ4_XS39.7 GB 42.9 GB
Too big Get file →
Q4_040.9 GB 44.1 GB
Too big Get file →
IQ4_NL41.3 GB 44.5 GB
Too big Get file →
Q4_K_S44.2 GB 47.4 GB
Too big Get file →
Q4_145.5 GB 48.6 GB
Too big Get file →
Q4_K_M47.7 GB 50.9 GB
Too big Get file →
Q5_050.0 GB 53.2 GB
Too big Get file →
Q5_154.5 GB (2 parts) 57.7 GB
Too big Get file →
Q5_K_M54.6 GB (2 parts) 57.8 GB
Too big Get file →
Q5_K_L55.2 GB (2 parts) 58.4 GB
Too big Get file →
Q6_K_M64.3 GB (2 parts) 67.5 GB
Too big Get file →
Q8_077.3 GB (2 parts) 80.4 GB
Too big Get file →
BF1679.6 GB (2 parts) 82.8 GB
Too big Get file →
F1679.6 GB (2 parts) 82.8 GB
Too big Get file →
BF16145.4 GB (4 parts) 148.6 GB
Too big Get file →
F16145.4 GB (4 parts) 148.6 GB
Too big Get file →

For 8 GB VRAM, use IQ1_M (24.19 GB file, 27.4 GB memory at 8 k context). For 16 GB VRAM, IQ2_S (27.51 GB file, 30.7 GB memory). For 24 GB VRAM, Q4_K_M (47.74 GB file, 50.9 GB memory). Higher‑bit quants require more memory and are not recommended for limited VRAM setups.

What GPU do you need for Qwen2.5-VL-72B-Instruct?

These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.

  • 4 GB VRAM: no quant fits at an 8K context.
  • 6 GB VRAM: no quant fits at an 8K context.
  • 8 GB VRAM: no quant fits at an 8K context.
  • 12 GB VRAM: no quant fits at an 8K context.
  • 16 GB VRAM: no quant fits at an 8K context.
  • 24 GB VRAM: no quant fits at an 8K context.
  • 32 GB VRAM: the best quant that fits is IQ2_S (28.3 GB file, about 31.4 GB in use).
  • 48 GB VRAM: the best quant that fits is Q4_K_S (44.2 GB file, about 47.4 GB in use).

Which Qwen2.5-VL-72B-Instruct quant should you choose?

Q4_K_M is the usual starting point for Qwen2.5-VL-72B-Instruct: it balances file size and quality well. Here it is a 47.7 GB file.

The Q8_0 file is 77.3 GB, about 1.6 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.

If memory is tight, the smallest file is IQ1_M at 24.2 GB, with the quality trade-offs that come with fewer bits. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architectureqwen2vl
Parameters73B
Layers80
Trained context128,000 tokens
KV cache (fp16)320 KB per token

Source

Quantized byMungert
FamilyQwen
Licenseother
Downloads (HF)1,204
Last updated2025-09-24

Frequently asked questions about Qwen2.5-VL-72B-Instruct

How much VRAM does Qwen2.5-VL-72B-Instruct need?

About 50.9 GB at Q4_K_M with an 8K context window. That is the 47.7 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.

Can I run Qwen2.5-VL-72B-Instruct on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 27.4 GB, so you would need a larger GPU or CPU offloading.

What is the file size of Qwen2.5-VL-72B-Instruct GGUF?

It depends on the quant: IQ1_M is 24.2 GB, IQ2_XXS is 25.9 GB, IQ2_XS is 27.5 GB, IQ2_S is 28.3 GB, IQ2_M is 29.7 GB, Q2_K_S is 29.9 GB and 22 more.

What context length does Qwen2.5-VL-72B-Instruct support?

The GGUF header lists a trained context length of 128,000 tokens (125K). The KV cache grows with the context you use: about 2.7 GB at 8K and 10.7 GB at 32K.

Where can I download Qwen2.5-VL-72B-Instruct GGUF?

The files are on Hugging Face in the Mungert/Qwen2.5-VL-72B-Instruct-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

What license does Qwen2.5-VL-72B-Instruct use?

Hugging Face lists the license as other. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.

Can I run Qwen2.5-VL-72B-Instruct on 8 GB VRAM?

Only the IQ1_M quantization can be attempted, but memory usage will exceed 8 GB; it is not recommended.

Does the model support vision tasks?

Yes, it can process images and videos, generate structured outputs, and localize objects.

What is the context length of the model?

The trained context length is 128000 tokens.

Other GGUFs of Qwen2.5-VL-72B-Instruct

Similar models

Browse all →

More from Mungert

All by Mungert →

More Qwen models

All Qwen →

Other large models

All large models →