[whatmodel.sh](../../README.md) / DeepSeek V4 Flash Vision (Exp)
# DeepSeek V4 Flash Vision (Exp)
DeepSeek · released 2026-08 · 304.6B params · mit
**score 73.6** ±2.3 · #45 of 732 · from 8 benchmark results across 2 sources
Score within the [LMArena (text)](https://whatmodel.sh/?set=lmarena-text&score=set) comparison set only.
## What it's good at
● marks DeepSeek V4 Flash Vision (Exp) among every model on that task or test, worst to best.
task vs the others rank counts tests
agent tasks ▁▁▃▆█▄▅▅▄●▂▁ #41 of 313 58% 4
coding ▁▁▁▃▅▇█▇▅●▃▁ #76 of 469* 12% 1
vision & documents ▁▁▂▂●▃▄▂▆▇█▄ #127 of 153* 1.0% 1
agent tasks · #41 of 313 · 58% of its score
Beats [GPT-5.6 Luna](gpt-5-6-luna.md), [Gemini 3.7 Flash](gemini-3-7-flash.md), [Muse Spark Open](muse-spark-open.md) here, though they rank higher overall. Trails [Grok 4.7](grok-4-7.md), [Claude Sonnet 5](claude-sonnet-5.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
Terminal-Bench 2.1 (maker) 83.9% ▁·▁·▁▁▄▃▆█●█ beats 49/63 hard · avg 25% 26% sharp · small run
DeepSWE v1.1 (resolved) (maker) 59.3% ▂····▂▁▁●███ beats 9/31 hard · avg 20% 12% at its level · small run
Agents' Last Exam (metric/harness… 27.3% ▂·▂·●▄▂·▂▂▄▄ beats 4/14 very hard · avg 7.4% 11% at its level · small run
Toolathlon-Verified (maker) 75.9% ▂·▂▂▅▃···▂●█ beats 12/18 medium · avg 36% 8.5% blunt · at its level · small run
coding · #76 of 469 · 12% of its score
Trails [Llama 3.1 Nemotron Ultra 253B v1](llama-3-1-nemotron-ultra-253b.md), [MiMo-V2.5 (310B-A15B)](mimo-v2-5.md), [MiMo-V2.5-Pro (1T-A42B)](mimo-v2-5-pro.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
DeepSWE v1.1 (resolved) (maker) 59.3% ▂····▂▁▁●███ beats 9/31 hard · avg 20% 12% at its level · small run
vision & documents · #127 of 153 · 1.0% of its score
Trails [Qwen3.5 2B](qwen3-5-2b.md), [Qwen3-VL 2B Thinking](qwen3-vl-2b-thinking.md), [Qwen2.5-VL 7B Instruct](qwen2-5-vl-7b.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
PureDocBench document parsing (Av… 50.1% ●·▁·▂▂▅█·▁▂▂ beats 0/22 medium · avg 62% 1.0% blunt · at its level
other tests (3)
test result vs the others in its run difficulty counts why
NL2Repo (maker) 57.7% ▂▂▂·▃▃▅█▃▂●▃ beats 18/22 very hard · avg 14% 17% at its level · small run
CyberGym (maker) 75.3% ▄···▄··▄●▄█▄ beats 3/8 hard · avg 34% 12% at its level · small run
AutomationBench (Public) (maker) 25.7% ●····▄····▄▄ beats 1/4 very hard · avg 6.9% 10% at its level · small run
source results share of the score
deepseek-ai/DeepSeek-V4-Flash-Vision-Ex… 7 97% ▇▇▇▇▇▇▇▇▇▇▇▇
How Far Is Document Parsing from Solved… 1 1% ▇
prior: pull toward the middle 2% ▇
weighted average = 73.6
Licences: deepseek-ai/DeepSeek-V4-Flash-Vision-Exp model card ([MIT](https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp/blob/6821d6ad3681a4b137b066b76094fa82ebd0a380/README.md)) · How Far Is Document Parsing from Solved? PureDocBench: A Source-Traceable Benchmark across Clean ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/))
## How to run it
### llama.cpp (recommended)
1. Install llama.cpp. Homebrew and winget installs name the commands llama-cli and llama-server instead. [Other ways to install](https://github.com/ggml-org/llama.cpp/blob/master/docs/install.md)
```sh
# macOS / Linux
curl -LsSf https://llama.app/install.sh | sh
# Windows
irm https://llama.app/install.ps1 | iex
```
2. Run DeepSeek V4 Flash Vision (Exp) (ud-q4_k_xl, needs ~156.03 GB, from [huggingface.co/unsloth](https://huggingface.co/unsloth/DeepSeek-V4-Flash-Vision-Exp-GGUF/tree/87ce15810931a270f165e5e50439cd446f6b4ba5))
```sh
llama cli -hf unsloth/DeepSeek-V4-Flash-Vision-Exp-GGUF:UD-Q4_K_XL
```
3. Or start a local server and open the address it prints (http://localhost:9931)
```sh
llama serve -hf unsloth/DeepSeek-V4-Flash-Vision-Exp-GGUF:UD-Q4_K_XL
```
Other versions: ud-q2_k_xl (~97.76 GB) `llama cli -hf unsloth/DeepSeek-V4-Flash-Vision-Exp-GGUF:UD-Q2_K_XL`
### LM Studio
1. Install LM Studio. Open the app once before using the lms command. [Download it (macOS / Windows / Linux)](https://lmstudio.ai/download)
2. Run DeepSeek V4 Flash Vision (Exp) (ud-q4_k_xl, needs ~156.03 GB, from [huggingface.co/unsloth](https://huggingface.co/unsloth/DeepSeek-V4-Flash-Vision-Exp-GGUF/tree/87ce15810931a270f165e5e50439cd446f6b4ba5))
```sh
lms get https://huggingface.co/unsloth/DeepSeek-V4-Flash-Vision-Exp-GGUF@UD-Q4_K_XL -y
```
3. Then run lms ls to see its name, and lms chat followed by that name
Other versions: ud-q2_k_xl (~97.76 GB) `lms get https://huggingface.co/unsloth/DeepSeek-V4-Flash-Vision-Exp-GGUF@UD-Q2_K_XL -y`
### Other ways to run it
1. Run it as published (fp8, needs ~167.82 GB, NVIDIA only, from [huggingface.co/deepseek-ai](https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp/tree/6821d6ad3681a4b137b066b76094fa82ebd0a380))
```sh
sglang serve --model-path deepseek-ai/DeepSeek-V4-Flash-Vision-Exp --tp 4 --speculative-algorithm DSPARK
```
Model repository and documentation: [huggingface.co/deepseek-ai](https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp)
[interactive version](https://whatmodel.sh/model/deepseek-v4-flash-vision-exp) · [JSON](https://whatmodel.sh/api?format=json&all=1)