Qwen3.8 27B MagicQuant needs about 20.3 GB of memory at Q4_K_M (a 17.6 GB file) with an 8K context window, so it fits on a 24 GB GPU. This repository has 13 quants, from 8.3 GB (IQ2_XXS) to 29.0 GB (Q6_K). Longer context needs more: at 32K tokens the KV cache alone adds about 8.7 GB.
Qwen3.8-27B MagicQuant GGUF model with vision support
Qwen3.8-27B MagicQuant is a 27.3B parameter vision‑language model. It supports image and video understanding and is optimized for llama.cpp inference. The model is released under the apache-2.0 license.
Good at
- Vision‑language tasks
- Long‑horizon agentic planning
- Coding and professional work
Limits
- No built‑in vision tools beyond image/video input
- Requires 8k context for memory estimates
- No external tool integration
Best for
- Image captioning
- Video summarization
- Complex multi‑step reasoning
Who it’s for. Developers needing a dense vision‑language model for local inference. The model is suitable for applications that require high‑quality image and video understanding without external APIs.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| IQ2_XXS | 8.3 GB | 11.0 GB | Too big | Get file → | |
| IQ2_M | 10.8 GB | 13.5 GB | Too big | Get file → | |
| IQ2_M | 12.0 GB | 14.6 GB | Too big | Get file → | |
| IQ4_XS | 16.3 GB | 19.0 GB | Too big | Get file → | |
| Q4_K_M | 17.6 GB | 20.3 GB | Too big | Get file → | |
| Q5_K_S | 19.0 GB | 21.7 GB | Too big | Get file → | |
| Q5_K | 20.1 GB | 22.8 GB | Too big | Get file → | |
| Q5_K | 20.9 GB | 23.6 GB | Too big | Get file → | |
| Q5_K | 22.0 GB | 24.7 GB | Too big | Get file → | |
| Q6_K | 23.2 GB | 25.9 GB | Too big | Get file → | |
| Q6_K | 25.9 GB | 28.6 GB | Too big | Get file → | |
| Q6_K | 27.3 GB | 29.9 GB | Too big | Get file → | |
| Q6_K | 29.0 GB | 31.7 GB | Too big | Get file → |
For 8‑GB VRAM, the IQ2_XXS quant (8.27 GB file, 11 GB memory at 8k context) is the lightest option. If 12‑GB VRAM is available, IQ2_M (10.77 GB file, 13.5 GB memory) offers a balance between size and quality. Larger quants such as Q4_K_M (17.62 GB file, 20.3 GB memory) require 20‑GB VRAM or more.
What GPU do you need for Qwen3.8 27B MagicQuant?
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: the best quant that fits is IQ2_XXS (8.3 GB file, about 11.0 GB in use).
- 16 GB VRAM: the best quant that fits is IQ2_M (12.0 GB file, about 14.6 GB in use).
- 24 GB VRAM: the best quant that fits is Q5_K (20.9 GB file, about 23.6 GB in use).
- 32 GB VRAM: the best quant that fits is Q6_K (29.0 GB file, about 31.7 GB in use).
Which Qwen3.8 27B MagicQuant quant should you choose?
Q4_K_M is the usual starting point for Qwen3.8 27B MagicQuant: it balances file size and quality well. Here it is a 17.6 GB file.
If memory is tight, the smallest file is IQ2_XXS at 8.3 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
Source
Frequently asked questions about Qwen3.8 27B MagicQuant
How much VRAM does Qwen3.8 27B MagicQuant need?
About 20.3 GB at Q4_K_M with an 8K context window. That is the 17.6 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 MagicQuant on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 11.0 GB, so you would need a larger GPU or CPU offloading.
What is the file size of Qwen3.8 27B MagicQuant GGUF?
It depends on the quant: IQ2_XXS is 8.3 GB, IQ2_M is 10.8 GB, IQ2_M is 12.0 GB, IQ4_XS is 16.3 GB, Q4_K_M is 17.6 GB, Q5_K_S is 19.0 GB and 7 more.
What context length does Qwen3.8 27B MagicQuant 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 MagicQuant GGUF?
The files are on Hugging Face in the magiccodingman/Qwen3.8-27B-MagicQuant-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 MagicQuant 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 MagicQuant on 8 GB VRAM?
The smallest quant, IQ2_XXS, needs 11 GB memory at 8k context, so 8 GB VRAM is insufficient. You would need at least 11 GB VRAM to run it.
What is the memory requirement for the Q4_K_M quant?
The Q4_K_M quant is 17.62 GB file size and requires 20.3 GB memory at 8k context.
Does this model support vision tasks?
Yes, the model has native vision‑language support for images and videos.