[whatmodel.sh](../../README.md) / DeepSeek V4 Pro (0813)
# DeepSeek V4 Pro (0813)
DeepSeek · released 2026-08 · 1600B params (49B active) · mit
**score 74.7** ±0.6 · #37 of 732 · from 32 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 Pro (0813) among every model on that task or test, worst to best.
task vs the others rank counts tests
agent tasks ▁▁▃▆█▄▅▅▄●▂▁ #40 of 313 24% 10
coding ▁▁▁▃▅▇█▇▅▄●▁ #32 of 469 13% 4
chat & writing ▁▁▁▃▆▄▆█▆▇●▃ #40 of 402* 3.8% 1
math ▁▁▃▅█▆▄█●▂▁▁ #38 of 431 34% 4
web search █●·········█ #2 of 3* 1.3% 1
agent tasks · #40 of 313 · 24% of its score
Beats [GPT-5.6 Luna](gpt-5-6-luna.md), [Gemini 3.7 Flash](gemini-3-7-flash.md) here, though they rank higher overall. Trails [Grok 4.7](grok-4-7.md), [Claude Sonnet 5](claude-sonnet-5.md), [Grok 4.5](grok-4-5.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
Arena Agent (October 2026 aggrega… -0.0030 IPS ▁▁▁▃▃●█▁▂▂▁▁ beats 24/59 avg -0.379 IPS 10% sharp
GameLogicBench solve rate (72 God… 41.7% █▄·█·▄·█●▄·▄ beats 8/11 very hard · avg 5.1% 2.7% at its level
Terminal-Bench 4.0 (maker) 12.4% ▁·▁··▁●▆▅██▃ beats 4/33 very hard · avg 0.4% 2.6% sharp · small run
Terminal-Bench 2.1 (maker) 87.9% ▁·▁·▁▁▄▃▆█▅● beats 58/63 hard · avg 29% 2.4% sharp · small run
Terminal-Bench Hard 51.4% ▂▅▆▂▂▆▂▂█▅▅● beats 27/27 very hard · avg 4.6% 2.2% sharp · at its level · small run
Humanity's Last Exam with tools (… 60.0% ▁·▁▂▂▁▄▄▇█●▅ beats 53/62 hard · avg 16% 1.3% at its level · small run
DeepSWE v1.1 (resolved) (maker) 62.7% ▂····▂▁▁●███ beats 10/31 hard · avg 23% 0.9% at its level · small run
Agents' Last Exam (pass@1, offici… 25.7% ●··········▄ beats 1/2 very hard · avg 7.0% 0.9% at its level · small run
Toolathlon-Verified (maker) 74.1% ▂·▂▂▅▃···▂●█ beats 10/18 hard · avg 35% 0.6% blunt · at its level · small run
Terminal-Bench 3.0 (maker) 11.8% █··●·······█ beats 1/2 very hard · avg 2.5% 0.5% noisy · small run
coding · #32 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 [GLM-5.2](glm-5-2.md), [GLM-5.3-Flash](glm-5-3-flash.md), [Grok 4.7](grok-4-7.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
Arena WebDev (Code Arena) 1581.9 ▁▁▃▂▅█▃▆●▅▁▂ beats 94/129 avg 981 6.8% sharp
GameLogicBench solve rate (72 God… 41.7% █▄·█·▄·█●▄·▄ beats 8/11 very hard · avg 5.1% 2.7% at its level
Luau Leetcode fully solved (Torpe… 82.2% ▁▁▁▁·▁▂▃▅▃●▇ beats 37/49 hard · avg 17% 2.6% sharp
DeepSWE v1.1 (resolved) (maker) 62.7% ▂····▂▁▁●███ beats 10/31 hard · avg 23% 0.9% at its level · small run
chat & writing · #40 of 402 · 3.8% of its score
Beats [GPT-6 Sol](gpt-6-sol.md), [MiMo-V2.6-Flash (309B-A15B)](mimo-v2-6-flash.md), [Claude Haiku 5.5](claude-haiku-5-5.md) here, though they rank higher overall. Trails [MiMo-V2.5-Pro (1T-A42B)](mimo-v2-5-pro.md), [Grok 4.20 (0309 reasoning)](grok-4-20.md), [Claude Opus 4.5](claude-opus-4-5.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
LMArena (text) 1465.1 ▁▁▁▂▃▃▃▆▆▇●▃ beats 339/385 avg 1260 3.8% blunt
math · #38 of 431 · 34% of its score
Trails [GPT-5.2](gpt-5-2.md), [Qwen3.7 Max](qwen3-7-max.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
FrontierMath tier 4 v2 26.8% ▁▂▁▂●▅▂▃▂▁▁▁ beats 32/67 very hard · avg 0.1% 16% sharp · at its level
FrontierMath tiers 1–3 v2 64.6% ▂··▁▁▃▄▄●▆▄▃ beats 50/86 very hard · avg 1.2% 16% sharp · at its level
MathArena Apex (pass@1) (maker) 65.3% ▄····▄···▄●▄ beats 3/5 hard · avg 17% 1.4% at its level · small run
OTIS mock AIME 2024–2025 98.6% ▃▃▄▃▃▃▇▇█▃●▅ beats 177/197 very hard · avg 7.2% 0.5% sharp · maxed out · noisy
web search · #2 of 3 · 1.3% of its score
test result vs the others in its run difficulty counts why
Humanity's Last Exam with tools (… 60.0% ▁·▁▂▂▁▄▄▇█●▅ beats 53/62 hard · avg 16% 1.3% at its level · small run
other tests (14)
test result vs the others in its run difficulty counts why
GPQA Diamond 91.7% ▁▁▁▁▂▄▆▄▅▇●▅ beats 187/211 hard · avg 38% 6.4% sharp · small run · several runs
Mystery Game Puzzles (Epoch) 43.0% █··▂▃▇▄█●▄▂▁ beats 63/77 very hard · avg 9.4% 6.2% sharp · noisy
Chess Puzzles (Epoch) 47.0% █▁▁▁▁▂▃▃▃●▁▁ beats 127/137 very hard · avg 5.8% 3.6% sharp · at its level · noisy
Humanity's Last Exam (maker) 42.7% ▁▁▁▄▆▆▇██●▁▃ beats 100/111 very hard · avg 5.4% 2.0% at its level · small run
NL2Repo (maker) 61.5% ▂▂▂·▃▃▅█▃▂●▃ beats 20/22 very hard · avg 15% 1.5% at its level · small run
SimpleQA Verified 52.9% ▁▁▃▂▃▃▆▆●▄▂▅ beats 59/81 very hard · avg 8.0% 1.5% at its level
Icelandic LLM Leaderboard average… 51.2% ▁▁▃▂▂▄▁▃●▄▄█ beats 20/38 very hard · avg 14% 1.4% at its level
SEC-Bench Pro (pass@1) (maker) 56.4% ▄·▄·▄·●▄···▄ beats 3/6 very hard · avg 8.7% 1.1% at its level · small run
CyberGym (maker) 83.3% ▄···▄··▄█▄●▄ beats 7/8 medium · avg 40% 1.0% small run
RobloxQA v2.0 multiple choice (To… 92.6% ▁▁▃▁█●▃▃▆▂▂▂ beats 23/49 easy · avg 77% 0.9% maxed out
AutomationBench (DeepSeek V4.1 ca… 43.2% █··●·······█ beats 1/2 very hard · avg 13% 0.8% at its level · small run
ExploitGym (pass@1) (maker) 5.4% ▄·▄··●·▄▄··▄ beats 2/6 very hard · avg 0.7% 0.8% too hard for it · noisy · small run
AutomationBench (Public) (maker) 31.8% █····●····▄▄ beats 2/4 very hard · avg 7.3% 0.7% at its level · small run
ProgramBench (Almost@1) (maker) 15.5% ●█·······█·█ beats 0/3 very hard · avg 5.2% 0.7% small run
source results share of the score
Epoch AI 7 47% ▇▇▇▇▇▇
deepseek-ai/DeepSeek-V4.1-Flash model c… 15 22% ▇▇▇
LMArena leaderboard dataset 3 20% ▇▇
Roblox LLM Leaderboard 2 3% ▇
GameLogicBench: Evaluating Coding Agent… 1 3% ▇
NVIDIA ModelOpt quantized-checkpoint ca… 1 2% ▇
Miðeind Icelandic LLM Leaderboard 1 1% ▇
deepseek-ai/DeepSeek-V4-Pro-0813 model … 2 1% ▇
prior: pull toward the middle 0% ▇
weighted average = 74.7
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)) · LMArena leaderboard dataset ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · 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/)) · NVIDIA ModelOpt quantized-checkpoint cards: baseline ([Apache-2.0 (Gemma-4-31B-IT-NVFP4 repo) / MIT (DeepSeek-V4-Pro-0813-NVFP4 repo)](https://www.apache.org/licenses/LICENSE-2.0)) · Miðeind Icelandic LLM Leaderboard ([CC BY 4.0](../../licenses/mideind-icelandic-llm-leaderboard-CC-BY-4.0.txt)) · 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) · $1.63/M blended · $1.3 in / $2.6 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-Pro.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-pro.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 Pro (0813) (ud-q4_k_xl, needs ~960.5 GB, from [huggingface.co/unsloth](https://huggingface.co/unsloth/DeepSeek-V4-Pro-0813-GGUF/tree/main))
```sh
llama cli -hf unsloth/DeepSeek-V4-Pro-0813-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-Pro-0813-GGUF:UD-Q4_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 Pro (0813) (ud-q4_k_xl, needs ~960.5 GB, from [huggingface.co/unsloth](https://huggingface.co/unsloth/DeepSeek-V4-Pro-0813-GGUF/tree/main))
```sh
lms get https://huggingface.co/unsloth/DeepSeek-V4-Pro-0813-GGUF@UD-Q4_K_XL -y
```
3. Then run lms ls to see its name, and lms chat followed by that name
Model repository and documentation: [huggingface.co/deepseek-ai](https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro-0813)
[interactive version](https://whatmodel.sh/model/deepseek-v4-pro) · [JSON](https://whatmodel.sh/api?format=json&all=1)