Skip to content
localllm
日本語

deepseek-ai

DeepSeek-V4-Flash-0731

Parameters
284.3B
Max context length
1,048,576
Architecture
deepseek4
License
Commercial OK
Base model
deepseek-ai/DeepSeek-V4-Flash-0731
License
mit
llama.cpp
b9840+

About DeepSeek-V4-Flash-0731

DeepSeek-V4-Flash-0731 is a GGUF model based on deepseek-ai/DeepSeek-V4-Flash-0731 on Hugging Face. GGUF files are published by unsloth/DeepSeek-V4-Flash-0731-GGUF. It has about 284.3B parameters, the deepseek4 architecture, a maximum context length of 1,048,576 tokens (values written in the GGUF).

There are 13 quantizations, with file sizes from 82.5 GB (UD-IQ1_S) to 162 GB (UD-Q8_K_XL) (actual sizes on Hugging Face).

Running UD-IQ1_S (82.5 GB) with an 8,192-token context needs an estimated 83.4 GB of memory, and an estimated 83.6 GB at 32,768 tokens. For 8,192 tokens, a GPU with 96 GB or more of VRAM, or a Mac with 96 GB or more of unified memory, is a rough guide (macOS caps how much memory the GPU can use, so leave some headroom).

The original model is licensed under MIT License, which allows commercial use (keep the license text and copyright notice when redistributing). Check the original license for the terms of use.

File size and estimated memory per quantization

  • Up to 8 GB
  • Up to 16 GB
  • Up to 24 GB
  • Up to 48 GB
  • Up to 128 GB
  • More

File sizes are the actual values on Hugging Face. Memory is estimated for each context length (KV cache f16). The formula is in How memory is calculated.

All memory figures are Estimated (KV cache f16). It should run on a setup with at least this much VRAM (or Mac unified memory). Click a value to see the devices and models that fit that amount (Find by GPU / Mac).

About the "quality guide" (how far a quantization drifts from the original)

For each quantization type, how far the output drifts from the original model, grouped into six levels from values llama.cpp measured on Llama-3-8B.

The value is mean KLD (the difference between the original and quantized model in next-token probabilities); smaller means closer to the original. Quantizations made with an imatrix (importance data) come out slightly smaller. Levels use the value without an imatrix (or with one, for types that only have that).

These values were not measured on this model. Larger models and MoE models tend to drift less with the same type. Distributor-specific mixed types (such as UD-Q4_K_XL) cannot be matched by type name alone, so they show "-".

Quantization Quality guide KLD (no imatrix) KLD (imatrix) Size (Llama-3-8B)
Q8_0 Nearly identical 0.0014 - 7.96 GiB
Q6_K Very small 0.0055 - 6.14 GiB
Q5_K_M Very small 0.0108 - 5.33 GiB
Q5_K_S Small 0.0166 - 5.21 GiB
Q5_1 Small 0.0180 - 5.65 GiB
Q5_0 Small 0.0222 - 5.21 GiB
Q4_K_M Small 0.0313 0.0282 4.58 GiB
Q4_K_S Medium 0.0431 0.0320 4.37 GiB
IQ4_NL Small - 0.0357 4.35 GiB
IQ4_XS Small - 0.0363 4.14 GiB
Q4_1 Medium 0.0717 - 4.78 GiB
Q4_0 Medium 0.0719 - 4.34 GiB
Q3_K_L Medium 0.0872 0.0731 4.03 GiB
Q3_K_M Medium 0.1019 0.0844 3.74 GiB
IQ3_M Medium - 0.1025 3.53 GiB
IQ3_S Medium - 0.1113 3.42 GiB
IQ3_XS Large - 0.1387 3.28 GiB
IQ3_XXS Large - 0.1836 3.05 GiB
Q3_K_S Large 0.2282 0.1998 3.41 GiB
IQ2_M Large - 0.3260 2.74 GiB
Q2_K Very large 0.4451 0.3322 2.96 GiB
Q2_K_S Very large - 0.4034 2.96 GiB
IQ2_S Very large - 0.4392 2.56 GiB
IQ2_XS Very large - 0.5463 2.43 GiB
IQ2_XXS Very large - 0.8120 2.24 GiB
IQ1_M Very large - 1.3931 2.01 GiB
IQ1_S Very large - 2.2546 1.88 GiB

Source: llama.cpp (MIT), measured on Llama-3-8B against FP16, Wikitext-2, revision f364eb6f. Original table · Reading and choosing quantizations

GGUF source

unsloth/DeepSeek-V4-Flash-0731-GGUF

Converted and published by a third party About this label

Last checked: Nov 11, 2026

Architecture
deepseek4
llama.cpp b9840 and later
Parameters
284.3B
Max context length
1,048,576
License
mit Commercial OK
File size and memory per quantization for DeepSeek-V4-Flash-0731 (unsloth/DeepSeek-V4-Flash-0731-GGUF)
Quantization File size Measured Quality guide 4,096 8,192 16,384 32,768 65,536 131,072 262,144 1,048,576 max
UD-IQ1_S (3 parts) 82.5 GB - 83.4 GB 83.4 GB 83.5 GB 83.6 GB 83.8 GB 84.3 GB 85.2 GB 90.6 GB
UD-IQ1_M (3 parts) 86.9 GB - 87.8 GB 87.8 GB 87.9 GB 88.0 GB 88.2 GB 88.7 GB 89.6 GB 95.0 GB
UD-IQ2_XXS (3 parts) 90.9 GB - 91.7 GB 91.8 GB 91.8 GB 91.9 GB 92.2 GB 92.6 GB 93.5 GB 98.9 GB
UD-IQ2_M (3 parts) 90.9 GB - 91.8 GB 91.8 GB 91.9 GB 92.0 GB 92.2 GB 92.7 GB 93.6 GB 99.0 GB
UD-Q2_K_XL (3 parts) 96.8 GB - 97.7 GB 97.7 GB 97.8 GB 97.9 GB 98.1 GB 98.6 GB 99.5 GB 105 GB
UD-IQ3_XXS (4 parts) 104 GB - 105 GB 105 GB 105 GB 105 GB 106 GB 106 GB 107 GB 112 GB
UD-IQ3_S (4 parts) 116 GB - 117 GB 117 GB 117 GB 117 GB 117 GB 118 GB 119 GB 124 GB
UD-Q3_K_M (4 parts) 128 GB - 129 GB 129 GB 129 GB 129 GB 129 GB 130 GB 131 GB 136 GB
UD-Q3_K_XL (4 parts) 128 GB - 129 GB 129 GB 129 GB 129 GB 130 GB 130 GB 131 GB 136 GB
UD-IQ4_NL (4 parts) 137 GB - 138 GB 138 GB 138 GB 138 GB 138 GB 138 GB 139 GB 145 GB
UD-IQ4_XS (4 parts) 137 GB - 138 GB 138 GB 138 GB 138 GB 138 GB 138 GB 139 GB 145 GB
UD-Q4_K_XL (5 parts) 155 GB - 156 GB 156 GB 156 GB 156 GB 156 GB 157 GB 158 GB 163 GB
UD-Q8_K_XL (5 parts) 162 GB - 163 GB 163 GB 163 GB 163 GB 163 GB 164 GB 165 GB 170 GB 
Ollama  ollama run hf.co/unsloth/DeepSeek-V4-Flash-0731-GGUF:UD-IQ1_S
llama.cpp  llama-server -hf unsloth/DeepSeek-V4-Flash-0731-GGUF:UD-IQ1_S

Third-party evaluations

Values from other organizations for the original model before quantization. Not this site's own rating. Values are for the original model before quantization. GGUF quantization (especially small quantizations) can lower performance.

Capabilities (Epoch AI)

154.4 152.5–156.4

DeepSeek V4 Flash 0731 · Values as of Jul 31, 2026

Source: Epoch Capabilities Index (Epoch AI) (CC BY 4.0). Original leaderboard · See in ranking

License and commercial use

MIT License Commercial OK

Allows commercial use, modification and redistribution. When redistributing, keep the copyright notice and license text.

Based on the license on the original model's Hugging Face page. Original license text (checked Oct 18, 2026)

The summary states facts this site read from the original text and is not legal advice. Always check the original license and its latest terms before use. All licenses

How to run

Mainline llama.cpp can load this model's format (deepseek4) from b9840 (Jun 29, 2026) onward. Ollama and LM Studio bundle llama.cpp internally, so an older app fails to load it with errors such as "unknown model architecture". If that happens, update the app.

Example commands using UD-IQ1_S. To use another quantization, replace the name with one from the table above. How to run local LLMs

Ollama

Run it by pointing directly at the GGUF on Hugging Face.

ollama run hf.co/unsloth/DeepSeek-V4-Flash-0731-GGUF:UD-IQ1_S

llama.cpp

Pass the repository and quantization name to llama-server (or llama-cli) with -hf.

llama-server -hf unsloth/DeepSeek-V4-Flash-0731-GGUF:UD-IQ1_S

LM Studio

Paste unsloth/DeepSeek-V4-Flash-0731-GGUF into the search box, download it, and choose the quantization from the file list.

Frequently asked questions

How much memory do I need to run DeepSeek-V4-Flash-0731 (UD-IQ1_S)?
The file size is 82.5 GB (measured). With a 8,192-token context, the estimated memory including the KV cache is 83.4 GB. The formula is published at https://localllm.az-arte.com/en/guide/memory-calculation.
Can DeepSeek-V4-Flash-0731 be used commercially?
The original model's license is MIT License (Commercial use allowed).Allows commercial use, modification and redistribution. When redistributing, keep the copyright notice and license text. Check the original license text before deciding.
Does DeepSeek-V4-Flash-0731 run in 8 GB, 12 GB, 16 GB or 24 GB of VRAM?
8 GB: no quantization fits; 12 GB: no quantization fits; 16 GB: no quantization fits; 24 GB: no quantization fits (estimated at 8,192 tokens with an f16 KV cache, files from the most-downloaded source).
What is the command to run DeepSeek-V4-Flash-0731 with Ollama?
Run "ollama run hf.co/unsloth/DeepSeek-V4-Flash-0731-GGUF:UD-IQ1_S" to use the GGUF on Hugging Face directly. To use another quantization, replace the name at the end.