QW

Qwen3.8 27B Omnimerge v6

QwenGGUF by ManniX-ITAUpdated 2026-09-16
Download on Hugging Face
Parameters27.3B
Context256K
Layers65
Architectureqwen35
Smallest quant9.6 GB
Quick answer

Qwen3.8 27B Omnimerge v6 needs about 19.5 GB of memory at Q4_K_M (a 16.8 GB file) with an 8K context window, so it fits on a 24 GB GPU. This repository has 20 quants, from 9.6 GB (IQ2_S) to 54.7 GB (F16). Longer context needs more: at 32K tokens the KV cache alone adds about 8.7 GB.

Qwen3.8-27B Omnimerge v6 – 27B parameter model with vision support

Qwen3.8-27B Omnimerge v6 is a 27‑billion‑parameter model built by arithmetic merging three Qwen3.6 fine‑tunes onto the Qwen3.8‑27B base. It retains the original vision projector and includes a MTP head for speculative decoding. The model is released under Apache‑2.0.

Good at

  • Vision tasks with preserved projector
  • MTP speculative decoding support
  • Large context up to 262,144 tokens

Limits

  • Requires 8‑k context memory estimates up to 57 GB for F16
  • No vision support in lower‑precision quantizations
  • MTP head only works with llama.cpp PR #22673

Best for

  • Vision‑enabled inference on high‑VRAM GPUs
  • Speculative decoding in research pipelines
  • Large‑context document analysis

Who it’s for. Developers needing a 27B model with vision and MTP support for research or production. The model is suitable for users with access to high‑VRAM GPUs and the latest llama.cpp build.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ2_S9.6 GB 12.3 GB
Too big Get file →
IQ2_M10.2 GB 12.9 GB
Too big Get file →
Q2_K11.0 GB 13.6 GB
Too big Get file →
IQ3_XXS11.4 GB 14.1 GB
Too big Get file →
IQ3_XS12.2 GB 14.9 GB
Too big Get file →
Q3_K_S12.3 GB 14.9 GB
Too big Get file →
IQ3_M12.8 GB 15.4 GB
Too big Get file →
Q3_K_M13.5 GB 16.2 GB
Too big Get file →
Q3_K_L14.6 GB 17.2 GB
Too big Get file →
Q3_K_XL14.6 GB 17.3 GB
Too big Get file →
IQ4_XS15.3 GB 18.0 GB
Too big Get file →
IQ4_NL16.0 GB 18.7 GB
Too big Get file →
Q4_K_M16.8 GB 19.5 GB
Too big Get file →
Q4_K_L17.8 GB 20.4 GB
Too big Get file →
Q5_K_M19.5 GB 22.2 GB
Too big Get file →
Q5_K_L20.3 GB 23.0 GB
Too big Get file →
Q6_K22.4 GB 25.1 GB
Too big Get file →
Q6_K_L23.0 GB 25.7 GB
Too big Get file →
Q8_029.0 GB 31.7 GB
Too big Get file →
F1654.7 GB 57.3 GB
Too big Get file →

For 8 GB VRAM, use IQ2_S (9.6 GB file, ~12.3 GB memory). For 16 GB VRAM, IQ4_XS (15.31 GB file, ~18 GB memory). For 24 GB VRAM, Q4_K_M (16.81 GB file, ~19.5 GB memory). Memory estimates are approximate.

What GPU do you need for Qwen3.8 27B Omnimerge v6?

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: the best quant that fits is IQ3_M (12.8 GB file, about 15.4 GB in use).
  • 24 GB VRAM: the best quant that fits is Q5_K_L (20.3 GB file, about 23.0 GB in use).
  • 32 GB VRAM: the best quant that fits is Q8_0 (29.0 GB file, about 31.7 GB in use).
  • 48 GB VRAM: the best quant that fits is Q8_0 (29.0 GB file, about 31.7 GB in use).

Which Qwen3.8 27B Omnimerge v6 quant should you choose?

Q4_K_M is the usual starting point for Qwen3.8 27B Omnimerge v6: it balances file size and quality well. Here it is a 16.8 GB file.

The Q8_0 file is 29.0 GB, about 1.7 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 IQ2_S at 9.6 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

Architectureqwen35
Parameters27.3B
Layers65
Trained context262,144 tokens
KV cache (fp16)260 KB per token

Source

Quantized byManniX-ITA
FamilyQwen
Licenseapache-2.0
Downloads (HF)15,351
Last updated2026-09-16

Frequently asked questions about Qwen3.8 27B Omnimerge v6

How much VRAM does Qwen3.8 27B Omnimerge v6 need?

About 19.5 GB at Q4_K_M with an 8K context window. That is the 16.8 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 Omnimerge v6 on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 12.3 GB, so you would need a larger GPU or CPU offloading.

What is the file size of Qwen3.8 27B Omnimerge v6 GGUF?

It depends on the quant: IQ2_S is 9.6 GB, IQ2_M is 10.2 GB, Q2_K is 11.0 GB, IQ3_XXS is 11.4 GB, IQ3_XS is 12.2 GB, Q3_K_S is 12.3 GB and 14 more.

What context length does Qwen3.8 27B Omnimerge v6 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 Omnimerge v6 GGUF?

The files are on Hugging Face in the ManniX-ITA/Qwen3.8-27B-Omnimerge-v6-MTP-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 Omnimerge v6 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 Qwen3.8-27B Omnimerge v6 on 8 GB VRAM?

Use the IQ2_S quantization; it requires about 12.3 GB of VRAM at 8k context, which is close to 8 GB but may need swapping.

Does the model support vision tasks?

Yes, the model includes the original Qwen3.8 vision projector and is vision‑capable.

What is MTP and how do I enable it?

MTP is a speculative decoding head. Enable it with llama.cpp PR #22673 and the --spec-type draft-mtp flag.

Other GGUFs of Qwen3.8-27B-Omnimerge-v6

Similar models

Browse all →

More from ManniX-ITA

All by ManniX-ITA →

More Qwen models

All Qwen →

Other medium models

All medium models →