Qwen3.8 27B FrostDFlash needs about 1.9 GB of memory at its main file (a 1.2 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 3 quants, from 1.0 GB to 1.5 GB. Longer context needs more: at 32K tokens the KV cache alone adds about 0.7 GB.
Qwen3.8 27B FrostDFlash – a dense draft model for speculative decoding
Qwen3.8 27B FrostDFlash is a ~1.9B‑parameter draft model that runs alongside Qwen3.8‑27B in llama.cpp speculative decoding. It uses custom precision‑gradient quantization and supports the DFlash2 drafter sidecards. The model is licensed under apache‑2.0.
Good at
- Speculative decoding with Qwen3.8‑27B
- Low‑precision draft generation
- Efficient sidecard integration
Limits
- No standalone LLM – requires target model
- No vision support in this draft
- Requires llama.cpp spec‑type draft‑dflash
Best for
- Speeding up inference for Qwen3.8‑27B
- Reducing VRAM usage during speculative decoding
- Testing draft model variants
Who it’s for. Developers building or testing speculative decoding pipelines with Qwen3.8‑27B. It is useful for those who need a lightweight draft model to reduce memory overhead during inference.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 1.0 GB | 1.7 GB | Fits | Get file → | |
| — | 1.2 GB | 1.9 GB | Fits | Get file → | |
| — | 1.5 GB | 2.1 GB | Fits | Get file → |
For a 8 GB VRAM budget, use the 1.05 GB file (est. 1.7 GB RAM at 8 k context). If more memory is available, the 1.23 GB file (est. 1.9 GB RAM) or the 1.47 GB file (est. 2.1 GB RAM) can be used. Estimates are approximate.
What GPU do you need for Qwen3.8 27B FrostDFlash?
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: the best quant that fits is the main file (1.5 GB file, about 2.1 GB in use).
Which Qwen3.8 27B FrostDFlash quant should you choose?
If memory is tight, the smallest file is the first one in the table at 1.0 GB, with the quality trade-offs that come with fewer bits.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Qwen3.8 27B FrostDFlash
How much VRAM does Qwen3.8 27B FrostDFlash need?
About 1.9 GB at its main file with an 8K context window. That is the 1.2 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 FrostDFlash on an 8 GB GPU?
Yes, with the smaller files: the 1.5 GB file needs about 2.1 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Qwen3.8 27B FrostDFlash GGUF?
It depends on the quant: file is 1.0 GB, file is 1.2 GB, file is 1.5 GB.
What context length does Qwen3.8 27B FrostDFlash support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 0.2 GB at 8K and 0.7 GB at 32K.
Where can I download Qwen3.8 27B FrostDFlash GGUF?
The files are on Hugging Face in the Plaguekind/Qwen3.8-27B-FrostDFlash-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 FrostDFlash 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 FrostDFlash on 8 GB VRAM?
Yes, the 1.05 GB file requires about 1.7 GB RAM at 8 k context, well within 8 GB VRAM.
Does this model support vision tasks?
No, this draft model does not include vision capabilities.
What license applies to this model?
The model is licensed under apache‑2.0.