Qwen3.6 35B DFlash needs about 0.9 GB of memory at Q4_K_M (a 0.3 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 3 quants, from 0.3 GB (Q4_K_M) to 1.0 GB (F16). Longer context needs more: at 32K tokens the KV cache alone adds about 0.5 GB.
Qwen3.6 35B DFlash – lightweight draft model for speculative decoding
Qwen3.6 35B DFlash is a DFlash draft model built on Qwen3.6-35B-A3B-DFlash. It is designed to work with a target model in a speculative decoding server, improving throughput while preserving output quality. The model supports reasoning and tools abilities.
Good at
- Speculative decoding
- Long‑context handling
- Parallel token proposals
Limits
- Requires a DFlash‑compatible inference server
- No standalone language model
- No vision support
Best for
- High‑throughput chat serving
- Long‑form content generation
- Batch inference with sliding‑window attention
Who it’s for. Developers building inference pipelines that need faster generation with Qwen3.6-35B. It is suitable for users who can run a DFlash‑compatible server and pair the drafter with a target model. The model is released under the MIT license.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q4_K_M | 0.3 GB | 0.9 GB | Fits | Get file → | |
| Q8_0 | 0.5 GB | 1.1 GB | Fits | Get file → | |
| F16 | 1.0 GB | 1.6 GB | Fits | Get file → |
For 8 GB VRAM, use Q4_K_M (0.29 GB file, ~0.9 GB memory at 8 k context). For 12 GB VRAM, Q8_0 (0.51 GB file, ~1.1 GB memory) is suitable. If more memory is available, F16 (0.96 GB file, ~1.6 GB memory) provides the highest precision.
What GPU do you need for Qwen3.6 35B DFlash?
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 F16 (1.0 GB file, about 1.6 GB in use).
Which Qwen3.6 35B DFlash quant should you choose?
Q4_K_M is the usual starting point for Qwen3.6 35B DFlash: it balances file size and quality well. Here it is a 0.3 GB file.
The Q8_0 file is 0.5 GB, about 1.8 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
Source
Frequently asked questions about Qwen3.6 35B DFlash
How much VRAM does Qwen3.6 35B DFlash need?
About 0.9 GB at Q4_K_M with an 8K context window. That is the 0.3 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 Qwen3.6 35B DFlash on an 8 GB GPU?
Yes, with F16: the 1.0 GB file needs about 1.6 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Qwen3.6 35B DFlash GGUF?
It depends on the quant: Q4_K_M is 0.3 GB, Q8_0 is 0.5 GB, F16 is 1.0 GB.
What context length does Qwen3.6 35B DFlash support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 0.1 GB at 8K and 0.5 GB at 32K.
Where can I download Qwen3.6 35B DFlash GGUF?
The files are on Hugging Face in the starskyzheng/Qwen3.6-35B-DFlash-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 Qwen3.6 35B DFlash use?
Hugging Face lists the license as mit. 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 Qwen3.6 35B DFlash on 8 GB VRAM?
Yes, the Q4_K_M quantization uses about 0.9 GB memory at 8 k context, fitting within 8 GB VRAM.
Does this model support vision tasks?
No, the model has no vision capabilities.
What server is required to use this model?
It must be paired with a DFlash‑compatible inference server, such as a fork of llama.cpp that includes DFlash support.