QW

Qwen3.8-27B

QwenGGUF by a4lgUpdated 2026-08-20
Download on Hugging Face
Parameters2.97B
Context256K
Layers65
Architectureqwen35
Smallest quant2.1 GB
Quick answer

Qwen3.8-27B needs about 4.8 GB of memory at Q4_K_M (a 2.1 GB file) with an 8K context window, so it fits on a 6 GB GPU. This repository has 4 quants, from 2.1 GB (Q4_K_M) to 5.9 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 8.7 GB.

Qwen3.8-27B MTP-only GGUF model for speculative decoding

Qwen3.8-27B is a 27B parameter dense model with a 262144 token context. It supports reasoning and tool usage. The GGUF subset contains only draft tensors for speculative decoding.

Good at

  • Reasoning
  • Tool execution
  • Long-horizon planning

Limits

  • Requires MTP tensors for full performance
  • No vision encoder in this subset
  • Large VRAM needed for higher quantizations

Best for

  • Speculative decoding acceleration
  • Research on draft model performance
  • Testing MTP settings

Who it’s for. Developers building inference pipelines with llama.cpp who need a lightweight draft model for speculative decoding. It is suitable for testing MTP configurations before grafting into full Qwen models.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q4_K_M2.1 GB 4.8 GB
Fits Get file →
Q6_K2.8 GB 5.4 GB
Fits Get file →
Q8_04.5 GB 7.2 GB
Fits Get file →
BF165.9 GB 8.6 GB
Too big Get file →

For 8 GB VRAM, use Q4_K_M (file size 2.12 GB, estimated 4.8 GB memory). For 12 GB VRAM, Q6_K (2.76 GB, 5.4 GB memory). For 16 GB VRAM, Q8_0 (4.5 GB, 7.2 GB memory).

What GPU do you need for Qwen3.8-27B?

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: the best quant that fits is Q6_K (2.8 GB file, about 5.4 GB in use).
  • 8 GB VRAM: the best quant that fits is Q8_0 (4.5 GB file, about 7.2 GB in use).
  • 12 GB VRAM: the best quant that fits is BF16 (5.9 GB file, about 8.6 GB in use).

Which Qwen3.8-27B quant should you choose?

Q4_K_M is the usual starting point for Qwen3.8-27B: it balances file size and quality well. Here it is a 2.1 GB file.

The Q8_0 file is 4.5 GB, about 2.1 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.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architectureqwen35
Parameters2.97B
Layers65
Trained context262,144 tokens
KV cache (fp16)260 KB per token

Source

Quantized bya4lg
FamilyQwen
Licenseapache-2.0
Downloads (HF)811
Last updated2026-08-20

Frequently asked questions about Qwen3.8-27B

How much VRAM does Qwen3.8-27B need?

About 4.8 GB at Q4_K_M with an 8K context window. That is the 2.1 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 Qwen3.8-27B on an 8 GB GPU?

Yes, with Q8_0: the 4.5 GB file needs about 7.2 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of Qwen3.8-27B GGUF?

It depends on the quant: Q4_K_M is 2.1 GB, Q6_K is 2.8 GB, Q8_0 is 4.5 GB, BF16 is 5.9 GB.

What context length does Qwen3.8-27B support?

The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 2.2 GB at 8K and 8.7 GB at 32K.

Where can I download Qwen3.8-27B GGUF?

The files are on Hugging Face in the a4lg/Qwen3.8-27B-MTP-ONLY-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 Qwen3.8-27B use?

Hugging Face lists the license as apache-2.0. 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 this model on 8 GB VRAM?

Yes, using the Q4_K_M quantization which requires about 4.8 GB of VRAM at 8k context.

Does this model support vision tasks?

The original Qwen3.8-27B has vision support, but this MTP-only GGUF subset does not include vision tensors.

What license applies to this model?

The model is licensed under apache-2.0.

Other GGUFs of Qwen3.8-27B

Similar models

Browse all →

More from a4lg

All by a4lg →

More Qwen models

All Qwen →

Other tiny models

All tiny models →