Gemini 4 Argon
Made by Google DeepMind, released 30 Sep 2026. Epoch AI lists its access as “Limited access”. Appears strongest at coding, knowledge, maths and writing among the models listed here.
In short
What it does well
- Coding: Top tier (2nd of 77 listed models)
- Knowledge: Top tier (1st of 43 listed models)
- Maths: Top tier (9th of 43 listed models)
- Writing: Top tier (2nd of 77 listed models)
Where it falls short
The four main ratings show no weakness. That is not evidence it has none.
What we don't know yet
- Fixing real software bugs: Not enough data
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
- Top tierHow often people preferred its answers to writing prompts. 1,527 on Arena's scale · Source for Gemini 4 Argon writing score (lmarena)
- Coding
- Top tierHow often people preferred its answers to coding questions. 1,538 on Arena's scale · Source for Gemini 4 Argon coding score (lmarena)
- Maths
- Top tierHow many very hard maths problems it solved. 86.0% · Source for Gemini 4 Argon maths score (epoch)
- Knowledge
- Top tierHow many short factual questions it answered correctly. 76.0% · Source for Gemini 4 Argon knowledge score (epoch)
Evidence completeness: Broad evidence, 12 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
No category stands out as clearly weaker than its overall position.
Show every category we can see
| Category | Position | Arena score | Votes | Source |
|---|---|---|---|---|
| Expert-level prompts Kind of request | 2nd of 77 | 1,549 95% range 1,532 to 1,566 | 1,219 | Source for Expert-level prompts |
| Hard prompts Kind of request | 1st of 77 | 1,542 95% range 1,534 to 1,549 | 7,454 | Source for Hard prompts |
| Long prompts Kind of request | 1st of 77 | 1,542 95% range 1,533 to 1,551 | 5,061 | Source for Long prompts |
| Longer conversations Kind of request | 1st of 77 | 1,542 95% range 1,526 to 1,557 | 1,664 | Source for Longer conversations |
| Chinese Language | 1st of 76 | 1,600 95% range 1,580 to 1,621 | 918 | Source for Chinese |
| English prompts Language | 1st of 77 | 1,529 95% range 1,520 to 1,538 | 4,532 | Source for English prompts |
| French Language | 1st of 72 | 1,550 95% range 1,522 to 1,578 | 517 | Source for French |
| Non-English prompts Language | 1st of 77 | 1,521 95% range 1,513 to 1,528 | 7,351 | Source for Non-English prompts |
| Russian Language | 1st of 77 | 1,535 95% range 1,520 to 1,551 | 1,632 | Source for Russian |
| Spanish Language | 4th of 70 | 1,505 95% range 1,479 to 1,532 | 493 | Source for Spanish |
| Entertainment, sport and media Topic | 1st of 77 | 1,520 95% range 1,508 to 1,532 | 3,157 | Source for Entertainment, sport and media |
| Law and government Topic | 1st of 76 | 1,537 95% range 1,517 to 1,558 | 938 | Source for Law and government |
| Medicine and health Topic | 1st of 75 | 1,544 95% range 1,522 to 1,565 | 802 | Source for Medicine and health |
| Science Topic | 1st of 77 | 1,546 95% range 1,532 to 1,561 | 1,929 | Source for Science |
| Software and IT Topic | 1st of 77 | 1,545 95% range 1,536 to 1,555 | 4,193 | 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
- Writing and creating content: good-enough alternative.
- Coding and building things: good-enough alternative.
- Business emails and everyday work: best-supported choice.
- Research and checking facts: best-supported choice.
Where an alternative might be more suitable
It is rated Top tier on every main test where it has a result.
What it is
- Maker
- Google DeepMind
- Release date (Epoch AI)
- 30 Sep 2026
- Access, as listed by Epoch AI
- Limited access
- Licence, as listed by Arena
- Proprietary
- Maker's country (Epoch AI)
- United States of America
- Last observation in our data
- 28 Oct 2026
- Data last changed
- 29 Oct 2026, 14:36 UTC
Epoch lists access as 'Limited access'.
We have not verified an official product page for this model, so we do not link to one.
How you can use it
Proprietary. The model's files are not published, so it is generally available only through its maker or partners the maker chooses.
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 “Limited access”. See models, apps and harnesses and how AI pricing works.
Price
Verified consumer plans that include this model
- Google AI Pro: GBP 18.99 per month. Maker's page, checked 28 Oct 2026. The page lists 'Access to Gemini 4 Argon' under this plan. Price as shown to a UK visitor on the check date.
- Google AI Ultra (from): GBP 79.99 per month. Maker's page, checked 28 Oct 2026. 'Starting at' price; the page lists 'Highest access to Gemini 4 Argon'. A higher-usage tier is listed at 189.99 GBP.
Prices change and can differ by country. Check the maker's page before you pay.
See Gemini plans and pricing (checked 28 Oct 2026). Page opened and read.
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) |
1,529 Arena score 95% range 1,522 to 1,535, 11,889 votes |
Top tier | effort: High, bradley-terry rating | Gemini 4 Argon (High) lmarena · text/latest |
28 Oct 2026 | Source for Gemini 4 Argon Arena text: overall preference score (lmarena) |
| Arena text: creative writing Arena 'text' leaderboard, category creative_writing |
1,527 Arena score 95% range 1,515 to 1,540, 2,720 votes |
Top tier | effort: High, bradley-terry rating | Gemini 4 Argon (High) lmarena · text/latest |
28 Oct 2026 | Source for Gemini 4 Argon Arena text: creative writing score (lmarena) |
| Arena text: coding Arena 'text' leaderboard, category coding |
1,538 Arena score 95% range 1,527 to 1,550, 2,813 votes |
Top tier | effort: High, bradley-terry rating | Gemini 4 Argon (High) lmarena · text/latest |
28 Oct 2026 | Source for Gemini 4 Argon Arena text: coding score (lmarena) |
| Arena text: maths Arena 'text' leaderboard, category math |
1,538 Arena score 95% range 1,516 to 1,561, 674 votes |
Top tier | effort: High, bradley-terry rating | Gemini 4 Argon (High) lmarena · text/latest |
28 Oct 2026 | Source for Gemini 4 Argon Arena text: maths score (lmarena) |
| Arena text: following instructions Arena 'text' leaderboard, category instruction_following |
1,533 Arena score 95% range 1,523 to 1,542, 4,236 votes |
Top tier | effort: High, bradley-terry rating | Gemini 4 Argon (High) lmarena · text/latest |
28 Oct 2026 | Source for Gemini 4 Argon Arena text: following instructions score (lmarena) |
| Arena text: business and management Arena 'text' leaderboard, category industry_business_and_management_and_financial_operations |
1,515 Arena score 95% range 1,502 to 1,529, 1,994 votes |
Top tier | effort: High, bradley-terry rating | Gemini 4 Argon (High) lmarena · text/latest |
28 Oct 2026 | Source for Gemini 4 Argon Arena text: business and management score (lmarena) |
| Arena WebDev: building web apps Arena 'webdev' leaderboard (Code Arena), category overall |
1,679 Arena score 95% range 1,668 to 1,691, 3,437 votes |
Top tier | effort: High, bradley-terry rating | Gemini 4 Argon (High) lmarena · webdev/latest |
28 Oct 2026 | Source for Gemini 4 Argon Arena WebDev: building web apps score (lmarena) |
| FrontierMath Tiers 1-3 (v2) FrontierMath Tiers 1-3 v2 (private set), benchmark released 2026-06-12 |
86.0% share of problems solved standard error ±2.1 points |
Top tier | effort: high, best score across scorers | gemini-4-argon_high epoch · frontiermath_tiers_1_3_v2.csv |
13 Oct 2026 | Source for Gemini 4 Argon FrontierMath Tiers 1-3 (v2) score (epoch) No public run log in the data file. |
| OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025, benchmark released 2024-12-19 |
100.0% share of problems solved standard error ±0.0 points |
Top tier | effort: high, best score across scorers | gemini-4-argon_high epoch · otis_mock_aime_2024_2025.csv |
13 Oct 2026 | Source for Gemini 4 Argon OTIS Mock AIME 2024-2025 score (epoch) No public run log in the data file. |
| GPQA Diamond GPQA Diamond, benchmark released 2023-11-20 |
96.4% share of questions correct standard error ±1.2 points |
Top tier | effort: high, best score across scorers | gemini-4-argon_high epoch · gpqa_diamond.csv |
13 Oct 2026 | Source for Gemini 4 Argon GPQA Diamond score (epoch) No public run log in the data file. |
| SimpleQA Verified SimpleQA Verified, benchmark released 2025-09-09 |
76.0% share of questions correct standard error ±1.4 points |
Top tier | effort: high, best score across scorers | gemini-4-argon_high epoch · simpleqa_verified.csv |
13 Oct 2026 | Source for Gemini 4 Argon SimpleQA Verified score (epoch) No public run log in the data file. |
| SWE-bench Verified (Epoch runs) | Not enough data | |||||
| FrontierCode (reported by Cognition) FrontierCode, benchmark released 2026-06-08, aggregation Mean@5 |
47.5% mean score over 5 attempts no uncertainty figure published |
Not rated (results use different tools) | effort: medium, tool: devin-cli, mean@5 | gemini-4-argon_medium epoch · frontiercode_external.csv |
not provided | Source for Gemini 4 Argon FrontierCode (reported by Cognition) score (epoch) · Run details for FrontierCode (reported by Cognition) |
Missing evidence
Not enough data for these tests:
- SWE-bench Verified (Epoch runs)
Missing results are not low scores and do not count against the 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.