DeepSeek V4 Flash 0731 Abliterated DS4 Quality128 needs about 8.7 GB of memory at its main file (a 7.3 GB file) with an 8K context window, so it fits on a 12 GB GPU. This repository has 3 quants, from 7.3 GB to 102.8 GB.
DeepSeek V4 Flash Abliterated DS4 Quality128 – a 284B parameter model
DeepSeek V4 Flash Abliterated DS4 Quality128 is a 284‑billion‑parameter model with a 1,048,576‑token context. It is released under the MIT license and is a rank‑1 abliterated derivative of DeepSeek‑V4‑Flash‑0731. The model is quantized with the DS4 Quality128 policy, preserving exact MXFP4 routed experts on selected layers.
Good at
- Large‑context generation
- High‑quality routed‑expert inference
- Stable on Metal and llama.cpp runtimes
Limits
- No vision support
- Requires 128 GB M1 Ultra for full resident quality
- Large memory footprint for full‑precision tensors
Best for
- Agentic chat with 1M‑token context
- Long‑form content creation
- Research on routed‑expert behavior
Who it’s for. Developers and researchers needing a high‑capacity DeepSeek model with preserved expert precision. Suitable for environments with ample GPU memory and support for DS4 or llama.cpp runtimes.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 7.3 GB | 8.7 GB | Too big | Get file → | |
| — | 7.3 GB | 8.7 GB | Too big | Get file → | |
| — | 102.8 GB | 115.7 GB | Too big | Get file → |
For 8 GB VRAM use the 7.3 GB GGUF files; they estimate 8.7 GB memory at 8 k context. For 128 GB M1 Ultra use the 102.83 GB GGUF to keep all MXFP4 experts in native form; it estimates 115.7 GB memory at 8 k context. The estimates are approximate.
What GPU do you need for DeepSeek V4 Flash 0731 Abliterated DS4 Quality128?
These figures add the file size, an fp16 KV cache at an 8K context window 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 the main file (7.3 GB file, about 8.7 GB in use).
- 16 GB VRAM: the best quant that fits is the main file (7.3 GB file, about 8.7 GB in use).
- 24 GB VRAM: the best quant that fits is the main file (7.3 GB file, about 8.7 GB in use).
- 32 GB VRAM: the best quant that fits is the main file (7.3 GB file, about 8.7 GB in use).
- 48 GB VRAM: the best quant that fits is the main file (7.3 GB file, about 8.7 GB in use).
Which DeepSeek V4 Flash 0731 Abliterated DS4 Quality128 quant should you choose?
If memory is tight, the smallest file is the first one in the table at 7.3 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
Frequently asked questions about DeepSeek V4 Flash 0731 Abliterated DS4 Quality128
How much VRAM does DeepSeek V4 Flash 0731 Abliterated DS4 Quality128 need?
About 8.7 GB at its main file with an 8K context window. That is the 7.3 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 DeepSeek V4 Flash 0731 Abliterated DS4 Quality128 on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 8.7 GB, so you would need a larger GPU or CPU offloading.
What is the file size of DeepSeek V4 Flash 0731 Abliterated DS4 Quality128 GGUF?
It depends on the quant: file is 7.3 GB, file is 7.3 GB, file is 102.8 GB.
What context length does DeepSeek V4 Flash 0731 Abliterated DS4 Quality128 support?
The GGUF header lists a trained context length of 1,048,576 tokens (1024K).
Where can I download DeepSeek V4 Flash 0731 Abliterated DS4 Quality128 GGUF?
The files are on Hugging Face in the apetersson/DeepSeek-V4-Flash-0731-Abliterated-DS4-Quality128 repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does DeepSeek V4 Flash 0731 Abliterated DS4 Quality128 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 DeepSeek V4 Flash Abliterated on 8 GB VRAM?
The 7.3 GB GGUF files estimate 8.7 GB memory at 8 k context, so 8 GB VRAM is insufficient for full inference.
Does this model support DSpark speculative decoding?
Yes, use the DSpark‑support companion GGUF for DS4 or the llamacpp‑DSpark‑support for a llama.cpp build.
What license applies to this model?
The model is released under the MIT license.