Qwen3.8 27B DSpark needs about 2.1 GB of memory at Q8_0 (a 1.5 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 2 quants, from 1.5 GB (Q8_0) to 2.7 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 0.7 GB.
Qwen3.8 27B DSpark – a speculative‑decoding drafter for Qwen3.8‑27B
Qwen3.8 27B DSpark is a draft model built on the Qwen family with a dflash architecture. It contains 1.36 billion parameters and supports a context length of 262,144 tokens. The model is released under the Apache‑2.0 license.
Good at
- Speculative decoding draft
- Large context support
- Qwen3.8 compatibility
Limits
- Experimental DSpark support
- High memory usage
- No vision capability
Best for
- Research on draft models
- Testing speculative decoding
- Evaluating Qwen3.8‑27B performance
Who it’s for. Developers and researchers who want to experiment with DSpark speculative decoding on the Qwen3.8‑27B architecture. The model is suitable for those who can run it on a system with sufficient VRAM and a llama.cpp build that includes PR #25173.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q8_0 | 1.5 GB | 2.1 GB | Fits | Get file → | |
| BF16 | 2.7 GB | 3.4 GB | Fits | Get file → |
For 8 GB VRAM, use the Q8_0 quant (file size 1.46 GB, estimated 2.1 GB memory at 8 k context). For 16 GB VRAM, the BF16 quant (file size 2.73 GB, estimated 3.4 GB memory at 8 k context) provides a balance between quality and memory usage. Memory estimates are approximate.
What GPU do you need for Qwen3.8 27B DSpark?
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 BF16 (2.7 GB file, about 3.4 GB in use).
Which Qwen3.8 27B DSpark quant should you choose?
Q8_0 (1.5 GB) is the middle option in this repository. Close to lossless. It costs roughly twice the memory of a 4-bit file, so it suits cases where you have plenty of VRAM.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Qwen3.8 27B DSpark
How much VRAM does Qwen3.8 27B DSpark need?
About 2.1 GB at Q8_0 with an 8K context window. That is the 1.5 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 Qwen3.8 27B DSpark on an 8 GB GPU?
Yes, with BF16: the 2.7 GB file needs about 3.4 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Qwen3.8 27B DSpark GGUF?
It depends on the quant: Q8_0 is 1.5 GB, BF16 is 2.7 GB.
What context length does Qwen3.8 27B DSpark 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 DSpark GGUF?
The files are on Hugging Face in the erlidev/Qwen3.8-27B-DSpark-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 DSpark 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 DSpark on 8 GB VRAM?
Yes, the Q8_0 quant requires about 2.1 GB of VRAM at 8 k context, which fits within an 8 GB budget.
Does this model support vision tasks?
No, the model has no vision capability.
Is the DSpark feature stable?
DSpark support is experimental and may not provide speedups; performance can be variable.