DeepSeek V4 Flash 0731 needs about 105.2 GB of memory at IQ2_M (a 104.0 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 14 quants, from 70.2 GB (IQ1_M) to 161.9 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 2.9 GB.
DeepSeek V4 Flash 0731 – 284B parameter LLM with speculative decoding
DeepSeek V4 Flash 0731 is a 284B parameter model based on the DeepSeek architecture. It supports speculative decoding and is released under the MIT license. The model is quantization‑aware‑trained and includes a variety of quantized GGUF files.
Good at
- Speculative decoding
- Agentic reasoning
- Large context support
Limits
- No vision support
- Requires updated chat template
- Large memory footprint for high‑precision quants
Best for
- Chatbots needing fast inference
- Coding assistants with reasoning loops
- Research prototyping on local hardware
Who it’s for. Developers and researchers who need a high‑capacity language model that can run locally with llama.cpp or vLLM. The model is suitable for users who can manage its memory requirements and are comfortable updating the chat template.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| IQ1_M | 70.2 GB (4 parts) | 71.5 GB | Too big | Get file → | |
| IQ1_M | 72.8 GB (4 parts) | 74.0 GB | Too big | Get file → | |
| IQ2_XXS | 78.5 GB (4 parts) | 79.8 GB | Too big | Get file → | |
| IQ2_XS | 85.1 GB (4 parts) | 86.4 GB | Too big | Get file → | |
| IQ2_S | 93.4 GB (4 parts) | 94.6 GB | Too big | Get file → | |
| IQ2_S | 96.8 GB (4 parts) | 98.0 GB | Too big | Get file → | |
| IQ2_M | 104.0 GB (4 parts) | 105.2 GB | Too big | Get file → | |
| IQ3_XXS | 108.1 GB (4 parts) | 109.3 GB | Too big | Get file → | |
| IQ3_XS | 118.2 GB (4 parts) | 119.4 GB | Too big | Get file → | |
| IQ3_S | 130.8 GB (4 parts) | 132.0 GB | Too big | Get file → | |
| IQ3_M | 135.8 GB (4 parts) | 137.0 GB | Too big | Get file → | |
| IQ3_M | 143.6 GB (4 parts) | 144.9 GB | Too big | Get file → | |
| MXFP4 | 154.5 GB (4 parts) | 155.8 GB | Too big | Get file → | |
| BF16 | 161.9 GB (4 parts) | 163.1 GB | Too big | Get file → |
For 8 GB VRAM, use IQ1_M (70.24 GB file, 79.2 GB memory at 8 k context). For 16 GB VRAM, IQ2_XS (85.14 GB file, 95.9 GB memory) or IQ2_S (93.4 GB file, 105.1 GB memory). For 32 GB VRAM, IQ3_M (135.79 GB file, 152.6 GB memory) or MXFP4 (154.55 GB file, 173.6 GB memory). Estimates are approximate.
What GPU do you need for DeepSeek V4 Flash 0731?
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: 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: no quant fits at an 8K context.
- 24 GB VRAM: no quant fits at an 8K context.
- 32 GB VRAM: no quant fits at an 8K context.
- 48 GB VRAM: no quant fits at an 8K context.
Which DeepSeek V4 Flash 0731 quant should you choose?
IQ2_M (104.0 GB) is the middle option in this repository. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.
If memory is tight, the smallest file is IQ1_M at 70.2 GB, with the quality trade-offs that come with fewer bits. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about DeepSeek V4 Flash 0731
How much VRAM does DeepSeek V4 Flash 0731 need?
About 105.2 GB at IQ2_M with an 8K context window. That is the 104.0 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 on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 71.5 GB, so you would need a larger GPU or CPU offloading.
What is the file size of DeepSeek V4 Flash 0731 GGUF?
It depends on the quant: IQ1_M is 70.2 GB, IQ1_M is 72.8 GB, IQ2_XXS is 78.5 GB, IQ2_XS is 85.1 GB, IQ2_S is 93.4 GB, IQ2_S is 96.8 GB and 8 more.
What context length does DeepSeek V4 Flash 0731 support?
The GGUF header lists a trained context length of 1,048,576 tokens (1024K). The KV cache grows with the context you use: about 0.7 GB at 8K and 2.9 GB at 32K.
Where can I download DeepSeek V4 Flash 0731 GGUF?
The files are on Hugging Face in the AtomicChat/DeepSeek-V4-Flash-0731-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 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 0731 on 8 GB VRAM?
Use the IQ1_M quant, which requires about 79.2 GB memory at 8 k context, so 8 GB VRAM is insufficient. A larger VRAM or a lower‑precision quant is needed.
Does the model support vision tasks?
No, the model has no vision capabilities.
What chat template should I use?
The embedded template is outdated; download the separate chat template file and pass it with --chat-template-file.