Gemini 4 Argon: benchmarks, price and rank
Gemini 4 Argon from Google is 4th of 123 rated AI models on Spezia's leaderboard, strongest on APEX-Agents and LMArena text. It costs $2 in, $10 out per million tokens.
- 4th of 123 rated models
- Rating 2,039
- Rated on 15 benchmarks
- Closed
- Released Sep 30, 2026
- 1 million tokens of context
Where it is strongest
- 2nd of 56 on APEX-Agents
- 2nd of 107 on LMArena text
- 3rd of 78 on SimpleQA Verified
- 3rd of 68 on SimpleBench
- 4th of 60 on Vending-Bench 2
Where it falls behind
- 15th of 18 on LifeSci Diamond
Price: $2 in, $10 out per million tokens.
Its headline results
The benchmarks that count most in the rating, with its place among the models measured the same way, and where each number came from.
- Remote Labor Index17.1%
Real freelance projects that people were paid to do, from design and video to architecture, games and data work. The model gets the same brief and files the freelancer got and has to hand back finished work.
5th of 17. Independent result from Epoch AI.
- APEX-Agents82.4%
Long tasks written by investment banking analysts, management consultants and corporate lawyers from their own work, done across files and apps in a realistic work setup.
2nd of 56. Independent result from Epoch AI.
- Terminal-Bench 4.057.4%
Hard jobs done in a real computer terminal: building and fixing software, setting up systems, and science and engineering problems.
6th of 17. The maker's own number, from Google, via LLM Stats.
- FrontierSWE55.0%
Open-ended technical projects that take hours to tens of hours, from making systems faster to building large programs and applied machine learning research. Each run gets a 20-hour budget.
7th of 21. Independent result from Epoch AI.
- Vending-Bench 2$13,718
The model runs a simulated vending machine business for a year on its own: finding suppliers, ordering stock, setting prices and paying the daily fees.
4th of 60. Independent result from Epoch AI.
- CursorBench52.9%
Ambiguous coding tasks that span many files, taken from real sessions in the Cursor code editor.
5th of 20. Independent result from Epoch AI.
- DeepSWE 1.177.9%
Software engineering tasks run in the same simple agent setup for every model.
3rd of 27. The maker's own number, from Google, via LLM Stats.
- SimpleBench91.7%
Multiple-choice puzzles about everyday situations, space, time and people, written to trip up models that pattern-match instead of think. Ordinary adults do well on them.
3rd of 68. Independent result from Epoch AI.
- SimpleQA Verified76%
Short factual questions with one checked answer, asked without web search.
3rd of 78. Independent result from Epoch AI.
- FrontierMath86.0%
Hundreds of new, unpublished math problems written by mathematicians, from hard undergraduate level up to research, with answers a computer can check.
13th of 76. Independent result from Epoch AI.
7 more results in its full profile.
Every result with its source, what it costs per task, what jumped with its release, side by side with any other model, and answers to your questions about when to use it.
Models ranked near it
How the rating works, and every benchmark it uses: the model leaderboard.