DeepSeek V4 Flash 0731 NVFP4 needs about 12.7 GB of memory at MXFP4 (a 10.9 GB file) with an 8K context window, so it fits on a 16 GB GPU. This repository has 2 quants, from 10.9 GB (MXFP4) to 165.0 GB.
DeepSeek V4 Flash 0731 NVFP4 GGUF model
DeepSeek V4 Flash 0731 NVFP4 is a GGUF conversion of the DeepSeek V4 model with NVIDIA NVFP4 weights. It contains 43 routed‑expert layers and a 1,048,576‑token context. The model is released under the MIT license.
Good at
- Large context handling
- NVFP4 calibrated experts
- High‑quality speculative decoding
Limits
- Requires 208 GiB VRAM or more for full GPU residency
- Main model size 165.04 GB, optional DSpark drafter 10.90 GB
- No vision capabilities
Best for
- Long‑form content generation
- Research prototyping with large context
- High‑throughput inference on powerful GPUs
Who it’s for. Engineers and researchers needing a DeepSeek V4 model with NVFP4 precision and large context. The model is suitable for deployment on systems with ample VRAM and the required llama.cpp runtime patch.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| MXFP4 | 10.9 GB | 12.7 GB | Too big | Get file → | |
| — | 165.0 GB (4 parts) | 185.3 GB | Too big | Get file → |
For systems with 8–12 GB VRAM, use the MXFP4 10.9 GB file; it estimates 12.7 GB memory at 8k context. For full‑size inference, download the 165.04 GB file split into four shards; it estimates 185.3 GB memory at 8k context. The optional DSpark drafter can be loaded separately to reduce VRAM usage.
What GPU do you need for DeepSeek V4 Flash 0731 NVFP4?
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: no quant fits at an 8K context.
- 16 GB VRAM: the best quant that fits is MXFP4 (10.9 GB file, about 12.7 GB in use).
- 24 GB VRAM: the best quant that fits is MXFP4 (10.9 GB file, about 12.7 GB in use).
- 32 GB VRAM: the best quant that fits is MXFP4 (10.9 GB file, about 12.7 GB in use).
- 48 GB VRAM: the best quant that fits is MXFP4 (10.9 GB file, about 12.7 GB in use).
Which DeepSeek V4 Flash 0731 NVFP4 quant should you choose?
MXFP4 (10.9 GB) is the middle option in this repository. Lower bit counts mean smaller files and more quality loss.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about DeepSeek V4 Flash 0731 NVFP4
How much VRAM does DeepSeek V4 Flash 0731 NVFP4 need?
About 12.7 GB at MXFP4 with an 8K context window. That is the 10.9 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 NVFP4 on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 12.7 GB, so you would need a larger GPU or CPU offloading.
What is the file size of DeepSeek V4 Flash 0731 NVFP4 GGUF?
It depends on the quant: MXFP4 is 10.9 GB, file is 165.0 GB.
Where can I download DeepSeek V4 Flash 0731 NVFP4 GGUF?
The files are on Hugging Face in the axiomofmind/DeepSeek-V4-Flash-0731-NVFP4-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 DeepSeek V4 Flash 0731 NVFP4 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 the main model on 8 GB VRAM?
No. The main model requires 208 GiB VRAM for full GPU residency, and the 165.04 GB file estimates 185.3 GB memory at 8k context.
What is the context length of this model?
The model advertises a 1,048,576‑token context. Longer contexts need more KV‑cache and compute memory.
Do I need a special runtime patch?
Yes. The NVFP4 layout requires the llama.cpp‑deepseek‑v4‑nvfp4‑runtime.patch applied to the specified llama.cpp commit.