QW

Qwen2-72B-Instruct

QwenGGUF by legraphistaUpdated 2024-06-07
Download on Hugging Face
Parameters72.7B
Context32K
Layers80
Architectureqwen2
Smallest quant22.7 GB
Quick answer

Qwen2-72B-Instruct needs about 38.7 GB of memory at IQ3_M (a 35.5 GB file) with an 8K context window, so it fits on a 48 GB GPU. This repository has 25 quants, from 22.7 GB (IQ1_S) to 145.4 GB. Longer context needs more: at 32K tokens the KV cache alone adds about 10.7 GB.

Qwen2-72B-Instruct – 72.7B parameter instruction‑tuned language model

Qwen2-72B-Instruct is a 72.7B parameter decoder model based on the Qwen2 architecture. It supports up to 131,072 token context length and is trained with supervised fine‑tuning and preference optimization. The model is available in multiple GGUF quantizations.

Good at

  • Large context handling
  • Multilingual text generation
  • Instruction following

Limits

  • No vision support
  • Requires large VRAM for higher quants
  • License is other

Best for

  • Chatbots needing long context
  • Research on long‑form text generation
  • Multilingual content creation

Who it’s for. Developers and researchers who need a large instruction‑tuned model with extended context. The model is suitable for deployment on systems with sufficient GPU memory and for use with llama.cpp or vLLM.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ1_S22.7 GB 25.9 GB
Too big Get file →
IQ1_M23.7 GB 26.9 GB
Too big Get file →
IQ2_XXS25.5 GB 28.7 GB
Too big Get file →
IQ2_XS27.1 GB 30.2 GB
Too big Get file →
IQ2_S27.9 GB 31.1 GB
Too big Get file →
IQ2_M29.3 GB 32.5 GB
Too big Get file →
Q2_K_S29.6 GB 32.8 GB
Too big Get file →
Q2_K29.8 GB 33.0 GB
Too big Get file →
IQ3_XXS31.8 GB 35.0 GB
Too big Get file →
IQ3_XS32.8 GB 36.0 GB
Too big Get file →
IQ3_S34.5 GB 37.7 GB
Too big Get file →
Q3_K_S34.5 GB 37.7 GB
Too big Get file →
IQ3_M35.5 GB 38.7 GB
Too big Get file →
Q3_K37.7 GB 40.9 GB
Too big Get file →
Q3_K_L39.5 GB 42.7 GB
Too big Get file →
IQ4_XS39.7 GB 42.9 GB
Too big Get file →
IQ4_NL41.3 GB 44.5 GB
Too big Get file →
Q4_K_S43.9 GB 47.1 GB
Too big Get file →
Q4_K47.4 GB (2 parts) 50.6 GB
Too big Get file →
Q5_K_S51.4 GB (2 parts) 54.6 GB
Too big Get file →
Q5_K54.4 GB (3 parts) 57.6 GB
Too big Get file →
Q6_K64.3 GB (3 parts) 67.5 GB
Too big Get file →
Q8_077.3 GB (4 parts) 80.4 GB
Too big Get file →
BF16145.4 GB (6 parts) 148.6 GB
Too big Get file →
—145.4 GB (6 parts) 148.6 GB
Too big Get file →

For 8 GB VRAM, use IQ1_S (22.69 GB file, 25.9 GB memory at 8 k context). For 16 GB VRAM, IQ2_S (27.94 GB file, 31.1 GB memory) is a good balance. For 24 GB VRAM, Q4_K (47.42 GB file, 50.6 GB memory) provides higher quality while staying within limits. Splits are required for Q4_K, Q5_K, Q6_K, Q8_0, BF16.

What GPU do you need for Qwen2-72B-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 (27.9 GB file, about 31.1 GB in use).
  • 48 GB VRAM: the best quant that fits is Q4_K_S (43.9 GB file, about 47.1 GB in use).

Which Qwen2-72B-Instruct quant should you choose?

IQ3_M (35.5 GB) is the middle option in this repository. A reasonable choice when a 4-bit file will not fit, mostly on larger models. Quality loss is noticeable but often acceptable.

If memory is tight, the smallest file is IQ1_S at 22.7 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

Architectureqwen2
Parameters72.7B
Layers80
Trained context32,768 tokens
KV cache (fp16)320 KB per token

Source

Quantized bylegraphista
FamilyQwen
Licenseother
Downloads (HF)706
Last updated2024-06-07

Frequently asked questions about Qwen2-72B-Instruct

How much VRAM does Qwen2-72B-Instruct need?

About 38.7 GB at IQ3_M with an 8K context window. That is the 35.5 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 Qwen2-72B-Instruct on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 25.9 GB, so you would need a larger GPU or CPU offloading.

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

It depends on the quant: IQ1_S is 22.7 GB, IQ1_M is 23.7 GB, IQ2_XXS is 25.5 GB, IQ2_XS is 27.1 GB, IQ2_S is 27.9 GB, IQ2_M is 29.3 GB and 19 more.

What context length does Qwen2-72B-Instruct support?

The GGUF header lists a trained context length of 32,768 tokens (32K). 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 Qwen2-72B-Instruct GGUF?

The files are on Hugging Face in the legraphista/Qwen2-72B-Instruct-IMat-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 Qwen2-72B-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 Qwen2-72B-Instruct on 8 GB VRAM?

Only the lowest quants such as IQ1_S or IQ1_M fit within 8 GB VRAM, but memory usage is approximate and may exceed the limit.

Does the model support vision tasks?

No, Qwen2-72B-Instruct is a text‑only model and does not support vision inputs.

What is the maximum context length?

The model supports up to 131,072 tokens, enabling very long text processing.

Other GGUFs of Qwen2-72B-Instruct

Similar models

Browse all →

More from legraphista

All by legraphista →

More Qwen models

All Qwen →

Other large models

All large models →