DeepSeek-V4-Pro
Made by DeepSeek, released 24 Apr 2026. Epoch AI lists its access as “Open weights (unrestricted)”. Appears solid at writing among the models listed here.
In short
What it does well
- Writing: Strong (31st of 77 listed models)
Where it falls short
- Maths: Behind the leaders (37th of 43 listed models)
- Coding: Middle (51st of 77 listed models)
- Knowledge: Middle (26th of 43 listed models)
What we don't know yet
Every test we use has a result.
The four ratings
- Top tier
- Strong
- Middle
- Behind the leaders
- Not enough data
Ratings compare only the 77 models listed here, in quarters, best first. “Not enough data” means no result yet. It is never a low score.
- Writing
- StrongHow often people preferred its answers to writing prompts. 1,446 on Arena's scale · Source for DeepSeek-V4-Pro writing score (lmarena)
- Coding
- MiddleHow often people preferred its answers to coding questions. 1,469 on Arena's scale · Source for DeepSeek-V4-Pro coding score (lmarena)
- Maths
- Behind the leadersHow many very hard maths problems it solved. 45.3% · Source for DeepSeek-V4-Pro maths score (epoch)
- Knowledge
- MiddleHow many short factual questions it answered correctly. 47.0% · Source for DeepSeek-V4-Pro knowledge score (epoch)
Evidence completeness: Broad evidence, 13 of 13 tests have a result.
Where it is relatively stronger or weaker
This compares each kind of request with the model's own overall position among the 77 listed models, using Arena's votes. It is about performance on different languages, kinds of request and topics, not about political or social bias.
Relatively stronger
No category stands out as clearly stronger than its overall position.
Relatively weaker
- Korean: 42nd of 54 on questions asked in Korean, against 42nd of 77 overall. Source for Korean
- Chinese: 57th of 76 on questions asked in Chinese, against 42nd of 77 overall. Source for Chinese
Show every category we can see
| Category | Position | Arena score | Votes | Source |
|---|---|---|---|---|
| Expert-level prompts Kind of request | 53rd of 77 | 1,463 95% range 1,454 to 1,471 | 5,922 | Source for Expert-level prompts |
| Hard prompts Kind of request | 46th of 77 | 1,461 95% range 1,456 to 1,466 | 38,501 | Source for Hard prompts |
| Long prompts Kind of request | 44th of 77 | 1,457 95% range 1,452 to 1,463 | 25,924 | Source for Long prompts |
| Longer conversations Kind of request | 34th of 77 | 1,467 95% range 1,460 to 1,474 | 10,354 | Source for Longer conversations |
| Chinese Language | 57th of 76 | 1,482 95% range 1,471 to 1,492 | 3,380 | Source for Chinese |
| English prompts Language | 40th of 77 | 1,459 95% range 1,454 to 1,464 | 25,284 | Source for English prompts |
| French Language | 37th of 72 | 1,473 95% range 1,458 to 1,487 | 2,210 | Source for French |
| German Language | 32nd of 55 | 1,457 95% range 1,437 to 1,476 | 995 | Source for German |
| Japanese Language | 25th of 54 | 1,445 95% range 1,422 to 1,469 | 725 | Source for Japanese |
| Korean Language | 42nd of 54 | 1,409 95% range 1,390 to 1,428 | 1,107 | Source for Korean |
| Non-English prompts Language | 46th of 77 | 1,439 95% range 1,434 to 1,444 | 32,288 | Source for Non-English prompts |
| Russian Language | 43rd of 77 | 1,453 95% range 1,445 to 1,461 | 6,314 | Source for Russian |
| Spanish Language | 38th of 70 | 1,456 95% range 1,441 to 1,471 | 1,728 | Source for Spanish |
| Entertainment, sport and media Topic | 40th of 77 | 1,431 95% range 1,424 to 1,437 | 12,804 | Source for Entertainment, sport and media |
| Law and government Topic | 39th of 76 | 1,464 95% range 1,455 to 1,474 | 4,487 | Source for Law and government |
| Medicine and health Topic | 35th of 75 | 1,466 95% range 1,456 to 1,476 | 4,060 | Source for Medicine and health |
| Science Topic | 38th of 77 | 1,470 95% range 1,463 to 1,477 | 9,171 | Source for Science |
| Software and IT Topic | 50th of 77 | 1,468 95% range 1,463 to 1,474 | 23,647 | Source for Software and IT |
Bias: what we can and can't see
Our sources do not measure political lean, stereotyping, flattery (agreeing with you too readily) or safety behaviour, so we cannot say how any model does on those. What we can show is above: where a model is relatively stronger or weaker across languages, kinds of request and topics. Also, people voting in Arena can favour answer style, such as length and formatting, and our numbers do not adjust for that. See what AI bias means and what these tests can tell you.
Best-supported use cases
This model is not among the options we recommend for any of the five tasks. That reflects the rules and the models listed, not a verdict on the model.
Where an alternative might be more suitable
- Writing and creating content: rated Strong on writing votes. Compare it with Claude Opus 5.5, the best-supported choice for this task.
- Coding and building things: rated Middle on coding votes. Compare it with Claude Opus 5.5, the best-supported choice for this task.
- Homework and maths: rated Behind the leaders on hard maths problems. Compare it with GPT-6.1 Sol, the best-supported choice for this task.
- Business emails and everyday work: rated Middle on business-task votes. Compare it with Gemini 4 Argon, the best-supported choice for this task.
- Research and checking facts: rated Middle on short factual questions. Compare it with Gemini 4 Argon, the best-supported choice for this task.
What it is
- Maker
- DeepSeek
- Release date (Epoch AI)
- 24 Apr 2026
- Access, as listed by Epoch AI
- Open weights (unrestricted)
- Licence, as listed by Arena
- MIT
- Maker's country (Epoch AI)
- China
- Last observation in our data
- 28 Oct 2026
- Data last changed
- 29 Oct 2026, 14:36 UTC
We have not verified an official product page for this model, so we do not link to one.
How you can use it
Open weights. Arena lists its licence as MIT. In plain terms, the model's files are published under a licence that lets others run it, so it can be offered by more than one company, sometimes at lower cost. That does not guarantee that any particular app is free. Open licences differ, and some restrict commercial use, so check the licence terms before building a business on it.
How you can reach it (an app, a website, a developer API or a third-party tool) and what that costs are not covered by our benchmark data. Where we could verify prices from the maker's own pages, they are shown under Price. Epoch AI lists access as “Open weights (unrestricted)”. See models, apps and harnesses and how AI pricing works.
Price
Price not independently verified
See DeepSeek API models and pricing (checked 29 Oct 2026). Page opened and read. It gives developer (API) prices per 1M tokens, shown on model pages. The chat app (chat.deepseek.com) goes straight to a sign-in page and we did not sign in, so we make no statement about whether it is free.
We never infer subscription prices from API prices. New to “tokens” and “per million”? See how AI pricing works.
Good to remember
- Benchmarks are samples of tasks. None proves performance in every situation.
- Arena scores come from people choosing between answers. That measures preference, not correctness.
- Ratings are relative to the 77 models listed here. A different set would change them.
- Sources test models at different settings. Each result shows the setting used, and we do not adjust it.
The evidence, for the curious
Show all 13 tests with settings and sources
| Test | Result | Rating | Setting used | Source identifier | Observed | Links |
|---|---|---|---|---|---|---|
| Arena text: overall preference Arena 'text' leaderboard (Bradley-Terry rating, no style control) |
1,451 Arena score 95% range 1,447 to 1,455, 57,572 votes |
Middle | effort: not stated, bradley-terry rating | DeepSeek-V4-Pro lmarena · text/latest |
28 Oct 2026 | Source for DeepSeek-V4-Pro Arena text: overall preference score (lmarena) |
| Arena text: creative writing Arena 'text' leaderboard, category creative_writing |
1,446 Arena score 95% range 1,439 to 1,454, 9,709 votes |
Strong | effort: not stated, bradley-terry rating | DeepSeek-V4-Pro lmarena · text/latest |
28 Oct 2026 | Source for DeepSeek-V4-Pro Arena text: creative writing score (lmarena) |
| Arena text: coding Arena 'text' leaderboard, category coding |
1,469 Arena score 95% range 1,463 to 1,475, 17,146 votes |
Middle | effort: not stated, bradley-terry rating | DeepSeek-V4-Pro lmarena · text/latest |
28 Oct 2026 | Source for DeepSeek-V4-Pro Arena text: coding score (lmarena) |
| Arena text: maths Arena 'text' leaderboard, category math |
1,442 Arena score 95% range 1,431 to 1,454, 3,061 votes |
Behind the leaders | effort: not stated, bradley-terry rating | DeepSeek-V4-Pro lmarena · text/latest |
28 Oct 2026 | Source for DeepSeek-V4-Pro Arena text: maths score (lmarena) |
| Arena text: following instructions Arena 'text' leaderboard, category instruction_following |
1,445 Arena score 95% range 1,440 to 1,451, 20,026 votes |
Middle | effort: not stated, bradley-terry rating | DeepSeek-V4-Pro lmarena · text/latest |
28 Oct 2026 | Source for DeepSeek-V4-Pro Arena text: following instructions score (lmarena) |
| Arena text: business and management Arena 'text' leaderboard, category industry_business_and_management_and_financial_operations |
1,443 Arena score 95% range 1,436 to 1,450, 11,378 votes |
Middle | effort: not stated, bradley-terry rating | DeepSeek-V4-Pro lmarena · text/latest |
28 Oct 2026 | Source for DeepSeek-V4-Pro Arena text: business and management score (lmarena) |
| Arena WebDev: building web apps Arena 'webdev' leaderboard (Code Arena), category overall |
1,447 Arena score 95% range 1,441 to 1,454, 13,931 votes |
Middle | effort: not stated, bradley-terry rating | DeepSeek-V4-Pro lmarena · webdev/latest |
28 Oct 2026 | Source for DeepSeek-V4-Pro Arena WebDev: building web apps score (lmarena) |
| FrontierMath Tiers 1-3 (v2) FrontierMath Tiers 1-3 v2 (private set), benchmark released 2026-06-12 |
45.3% share of problems solved standard error ±3.0 points |
Behind the leaders | effort: max, best score across scorers | deepseek-v4-pro_max epoch · frontiermath_tiers_1_3_v2.csv |
17 Jun 2026 | Source for DeepSeek-V4-Pro FrontierMath Tiers 1-3 (v2) score (epoch) · Run details for FrontierMath Tiers 1-3 (v2) |
| OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025, benchmark released 2024-12-19 |
96.7% share of problems solved standard error ±2.0 points |
Middle | effort: max, best score across scorers | deepseek-v4-pro_max epoch · otis_mock_aime_2024_2025.csv |
17 Jun 2026 | Source for DeepSeek-V4-Pro OTIS Mock AIME 2024-2025 score (epoch) No public run log in the data file. |
| GPQA Diamond GPQA Diamond, benchmark released 2023-11-20 |
89.6% share of questions correct standard error ±1.7 points |
Middle | effort: max, best score across scorers | deepseek-v4-pro_max epoch · gpqa_diamond.csv |
16 Jun 2026 | Source for DeepSeek-V4-Pro GPQA Diamond score (epoch) No public run log in the data file. |
| SimpleQA Verified SimpleQA Verified, benchmark released 2025-09-09 |
47.0% share of questions correct standard error ±1.6 points |
Middle | effort: max, best score across scorers | deepseek-v4-pro_max epoch · simpleqa_verified.csv |
27 Aug 2026 | Source for DeepSeek-V4-Pro SimpleQA Verified score (epoch) · Run details for SimpleQA Verified |
| SWE-bench Verified (Epoch runs) SWE-bench Verified, benchmark released 2024-08-13, as run by Epoch AI |
77.6% share of issues resolved standard error ±1.9 points |
Strong | effort: max, best score across scorers | deepseek-v4-pro_max epoch · swe_bench_verified.csv |
18 Jun 2026 | Source for DeepSeek-V4-Pro SWE-bench Verified (Epoch runs) score (epoch) · Run details for SWE-bench Verified (Epoch runs) |
| FrontierCode (reported by Cognition) FrontierCode, benchmark released 2026-06-08, aggregation Mean@5 |
17.6% mean score over 5 attempts no uncertainty figure published |
Not rated (results use different tools) | effort: none, tool: mini-swe-agent, mean@5 | deepseek-v4-pro_none epoch · frontiercode_external.csv |
not provided | Source for DeepSeek-V4-Pro FrontierCode (reported by Cognition) score (epoch) · Run details for FrontierCode (reported by Cognition) |
Missing evidence
Every test we use has a result for this model.
Where a source gives no public run log, we cannot show the run behind the score. See data sources.
Not sure this is the right one?
Tell us what you need help with and we'll narrow the options down.