[whatmodel.sh](../../README.md) / DeepSeek V4 Flash (0731)
# DeepSeek V4 Flash (0731)
DeepSeek · released 2026-07 · 284B params (13B active) · mit
**score 72.4** ±0.6 · #55 of 732 · from 41 benchmark results across 8 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 (0731) among every model on that task or test, worst to best.
task vs the others rank counts tests
agent tasks ▁▁▃▆█▄▅▅●▄▂▁ #50 of 313 15% 9
coding ▁▁▁▃▅▇█▇▅▄●▁ #38 of 469 13% 4
chat & writing ▁▁▁▃▆▄▆█▆▇●▃ #73 of 402* 3.6% 1
math ▁▁▃▅█▆▄█●▂▁▁ #43 of 431 32% 4
web search ●█·········█ #3 of 3* 1.2% 1
agent tasks · #50 of 313 · 15% of its score
Beats [Gemini 3.1 Pro (Preview)](gemini-3-1-pro.md), [Qwen3.8-Flash-Next](qwen3-8-flash-next.md), [Claude Opus 4.6](claude-opus-4-6.md) here, though they rank higher overall. Trails [Claude Sonnet 4.6](claude-sonnet-4-6.md), [GLM-5.2](glm-5-2.md), [GLM-5.3-Flash](glm-5-3-flash.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
Terminal-Bench 2.1 (maker) 82.7% ▁·▁·▁▁▄▃▆█●█ beats 47/63 hard · avg 29% 4.3% sharp · small run · several runs
GameLogicBench solve rate (72 God… 31.9% █▄·█·●·██▄·▄ beats 5/11 very hard · avg 5.1% 2.4% at its level
Terminal-Bench 4.0 (maker) 7.0% ▁·▁··●▆▆▅██▃ beats 2/33 very hard · avg 0.4% 2.1% sharp · small run
DeepSWE v1.1 (resolved) (maker) 54.4% ▂····▂▁●▆███ beats 5/31 hard · avg 23% 1.7% at its level · small run · several runs
Humanity's Last Exam with tools (… 51.5% ▁·▁▂▂▁▄▄▇●▄▅ beats 38/62 hard · avg 16% 1.2% at its level · small run
Toolathlon-Verified (maker) 70.3% ▂·▂▂▅▃···●██ beats 8/18 hard · avg 35% 1.2% blunt · at its level · small run · seve…
Agents' Last Exam (pass@1, offici… 25.2% ●··········▄ beats 0/2 very hard · avg 7.0% 0.8% at its level · small run
Agents' Last Exam (metric/harness… 25.2% ▂·▂·●▄▂·▂▂▄▄ beats 2/14 very hard · avg 7.4% 0.7% at its level · small run
Terminal-Bench 3.0 (maker) 7.6% ●··█·······█ beats 0/2 very hard · avg 2.5% 0.4% noisy · small run
coding · #38 of 469 · 13% of its score
Beats [GPT-5.5](gpt-5-5.md), [GPT-5.6 Terra](gpt-5-6-terra.md), [Claude Opus 4.8](claude-opus-4-8.md) here, though they rank higher overall. Trails [Qwen3.8 27B](qwen3-8-27b.md), [GLM-5.2](glm-5-2.md), [GLM-5.3-Flash](glm-5-3-flash.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
Arena WebDev (Code Arena) 1581.8 ▁▁▃▂▅█▃▆●▅▁▂ beats 93/129 avg 981 6.4% sharp
Luau Leetcode fully solved (Torpe… 77.9% ▁▁▁▁·▁▂▃▅▃●▇ beats 29/49 hard · avg 17% 2.8% sharp · at its level
GameLogicBench solve rate (72 God… 31.9% █▄·█·●·██▄·▄ beats 5/11 very hard · avg 5.1% 2.4% at its level
DeepSWE v1.1 (resolved) (maker) 54.4% ▂····▂▁●▆███ beats 5/31 hard · avg 23% 1.7% at its level · small run · several runs
chat & writing · #73 of 402 · 3.6% of its score
Beats [Claude Haiku 5.5](claude-haiku-5-5.md) here, though they rank higher overall. Trails [Gemma 4 26B-A4B](gemma-4-26b-a4b.md), [Nemotron Cascade 2 30B-A3B](nemotron-cascade-2-30b-a3b.md), [Gemma 4 31B](gemma-4-31b.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
LMArena (text) 1435.7 ▁▁▁▂▃▃▃▆▆▇●▃ beats 297/385 avg 1260 3.6% blunt
math · #43 of 431 · 32% of its score
Beats [Grok 4.7](grok-4-7.md), [Gemini 3.5 Flash](gemini-3-5-flash.md) here, though they rank higher overall. Trails [GPT-5.2](gpt-5-2.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
FrontierMath tiers 1–3 v2 57.5% ▂··▁▁▃▄▄●▆▄▃ beats 44/86 very hard · avg 1.2% 17% sharp · at its level
FrontierMath tier 4 v2 24.4% ▁▂▁▂●▅▂▃▂▁▁▁ beats 26/67 very hard · avg 0.1% 12% sharp · at its level
MathArena Apex (pass@1) (maker) 58.6% ▄····▄···●█▄ beats 2/5 hard · avg 17% 1.4% at its level · small run
OTIS mock AIME 2024–2025 94.4% ▃▃▄▃▃▃▇▇█●▂▅ beats 159/197 very hard · avg 7.2% 0.7% sharp · maxed out · noisy
web search · #3 of 3 · 1.2% of its score
test result vs the others in its run difficulty counts why
Humanity's Last Exam with tools (… 51.5% ▁·▁▂▂▁▄▄▇●▄▅ beats 38/62 hard · avg 16% 1.2% at its level · small run
other tests (18)
test result vs the others in its run difficulty counts why
GPQA Diamond 91.0% ▁▁▁▁▂▄▆▄▅▇●▅ beats 183/211 hard · avg 38% 7.2% sharp · small run · several runs
KMMLU-HARD (Daekeun Kim, zero-sho… 74.2% █···▁▁▁▂▃·●▂ beats 17/23 very hard · avg 29% 4.8% sharp · at its level
Mystery Game Puzzles (Epoch) 34.0% █··▂▃▇▄●▂▄▂▁ beats 53/77 very hard · avg 9.4% 4.5% sharp · noisy
Chess Puzzles (Epoch) 33.0% █▁▁▁▁▂▃▃●▂▁▁ beats 110/137 very hard · avg 5.8% 3.0% sharp · noisy
NL2Repo (maker) 54.2% ▂▂▂·▃▃▅█▃●▅▃ beats 17/22 very hard · avg 15% 2.6% at its level · small run · several runs
CLIcK Korean (Daekeun Kim) 90.6% ▆▄▄▄█▄▂·●█·▄ beats 17/23 medium · avg 56% 2.3% typical
Humanity's Last Exam (maker) 37.8% ▁▁▁▄▆▆▇██●▁▃ beats 93/111 very hard · avg 5.4% 1.9% at its level · small run
CyberGym (maker) 76.7% ▄···▄··▄●▄█▄ beats 4/8 medium · avg 40% 1.8% small run · several runs
KoBALT-700 (Daekeun Kim) 73.3% ●▄█·····▄··▄ beats 0/5 hard · avg 39% 1.8% at its level
HAE-RAE 1.0 (Daekeun Kim) 91.9% ▄▄▄▂▂█▂▄▂●▂▄ beats 17/22 medium · avg 61% 1.7% typical
KMMLU-Pro (Daekeun Kim) 79.2% ●·▄█·····▄·▄ beats 0/5 medium · avg 43% 1.6% typical
SimpleQA Verified 33.6% ▁▁▃▂▃▃●▆█▄▂▅ beats 25/81 very hard · avg 8.0% 1.4% at its level
AutomationBench (Public) (maker) 25.1% ●····▄····▄▄ beats 0/4 very hard · avg 6.9% 1.3% at its level · small run · several runs
MuSR Korean (Daekeun Kim) 81.2% █·····●██·██ beats 1/5 medium · avg 45% 1.1% typical
SEC-Bench Pro (pass@1) (maker) 30.9% ▄·●·▄·█▄···▄ beats 1/6 very hard · avg 8.7% 1.1% at its level · small run
RobloxQA v2.0 multiple choice (To… 91.8% ▁▁▃▁█●▃▃▆▂▂▂ beats 20/49 easy · avg 77% 1.0% maxed out
AutomationBench (DeepSeek V4.1 ca… 37.7% ●··█·······█ beats 0/2 very hard · avg 13% 0.8% at its level · small run
ExploitGym (pass@1) (maker) 1.8% ▄·●··█·▄▄··▄ beats 1/6 very hard · avg 0.7% 0.7% too hard for it · noisy · small run
source results share of the score
Epoch AI 7 42% ▇▇▇▇▇
deepseek-ai/DeepSeek-V4.1-Flash model c… 14 20% ▇▇
Daekeun Kim Korean LLM benchmark runs 6 13% ▇▇
LMArena leaderboard dataset 2 10% ▇
deepseek-ai/DeepSeek-V4-Flash-Vision-Ex… 7 6% ▇
Roblox LLM Leaderboard 2 4% ▇
GameLogicBench: Evaluating Coding Agent… 1 2% ▇
deepseek-ai/DeepSeek-V4-Pro-0813 model … 2 1% ▇
prior: pull toward the middle 0% ▇
weighted average = 72.4
Licences: Epoch AI ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · deepseek-ai/DeepSeek-V4.1-Flash model card ([MIT](../../licenses/deepseek-ai--DeepSeek-V4.1-Flash.txt)) · Daekeun Kim Korean LLM benchmark runs ([MIT](../../licenses/korean-personal-benchmark-runs-MIT.txt)) · LMArena leaderboard dataset ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · deepseek-ai/DeepSeek-V4-Flash-Vision-Exp model card ([MIT](https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Vision-Exp/blob/6821d6ad3681a4b137b066b76094fa82ebd0a380/README.md)) · Roblox LLM Leaderboard ([MIT](../../licenses/roblox-llm-leaderboard-MIT.txt)) · GameLogicBench: Evaluating Coding Agents on Runtime Game Logic with Tick-Level State Assertions ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · deepseek-ai/DeepSeek-V4-Pro-0813 model card ([MIT](../../licenses/deepseek-ai--DeepSeek-V4-Pro-0813.txt))
## How to run it
### Through an API
api (Deep Infra) · $0.11/M blended · $0.09 in / $0.18 out per 1M tokens · [deepinfra.com](https://deepinfra.com/models)
[price source](https://github.com/anomalyco/models.dev/blob/a668908cfe7073faa0363699486fc41a3ff52412/providers/deepinfra/models/deepseek-ai/DeepSeek-V4-Flash.toml): Standard base token rate from models.dev. Provider-specific tiers, long-context surcharges, caching and promotions may differ. · [context source](https://github.com/anomalyco/models.dev/blob/a668908cfe7073faa0363699486fc41a3ff52412/models/deepseek/deepseek-v4-flash.toml)
### 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 (0731) (ud-q4_k_xl, needs ~170.9 GB, from [huggingface.co/unsloth](https://huggingface.co/unsloth/DeepSeek-V4-Flash-0731-GGUF/tree/main))
```sh
llama cli -hf unsloth/DeepSeek-V4-Flash-0731-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-0731-GGUF:UD-Q4_K_XL
```
Other versions: ud-q2_k_xl (~85.7 GB) `llama cli -hf unsloth/DeepSeek-V4-Flash-0731-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 (0731) (ud-q4_k_xl, needs ~170.9 GB, from [huggingface.co/unsloth](https://huggingface.co/unsloth/DeepSeek-V4-Flash-0731-GGUF/tree/main))
```sh
lms get https://huggingface.co/unsloth/DeepSeek-V4-Flash-0731-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 (~85.7 GB) `lms get https://huggingface.co/unsloth/DeepSeek-V4-Flash-0731-GGUF@UD-Q2_K_XL -y`
Model repository and documentation: [huggingface.co/deepseek-ai](https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-0731)
[interactive version](https://whatmodel.sh/model/deepseek-v4-flash) · [JSON](https://whatmodel.sh/api?format=json&all=1)