Gemini 3.8 Flash: benchmarks, price and rank
Gemini 3.8 Flash from Google is 20th of 123 rated AI models on Spezia's leaderboard, strongest on LMArena text and GPQA Diamond. It costs $0.75 in, $3.75 out per million tokens.
- 20th of 123 rated models
- Rating 1,597
- Rated on 16 benchmarks
- Closed
- Released Sep 2, 2026
- 1 million tokens of context
- Knows the world to March 2026
Where it is strongest
- 8th of 107 on LMArena text
- 5th of 103 on GPQA Diamond
Where it falls behind
- 17th of 17 on Terminal-Bench 4.0
- 18th of 21 on FrontierSWE
- 17th of 20 on CursorBench
Price: $0.75 in, $3.75 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 Index5.8%
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.
9th of 17. Independent result from Epoch AI.
- APEX-Agents64.3%
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.
15th of 56. Independent result from Epoch AI.
- Terminal-Bench 4.019.1%
Hard jobs done in a real computer terminal: building and fixing software, setting up systems, and science and engineering problems.
17th of 17. The maker's own number, from Google, via LLM Stats.
- FrontierSWE19.6%
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.
18th of 21. Independent result from Epoch AI.
- Humanity's Last Exam44.5%
About 2,500 questions written by experts in more than a hundred fields to be as hard as they could make them, many needing a diagram or image.
Independent result from Epoch AI.
- Vending-Bench 2$5,094
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.
34th of 60. Independent result from Epoch AI.
- CursorBench39.6%
Ambiguous coding tasks that span many files, taken from real sessions in the Cursor code editor.
17th of 20. Independent result from Epoch AI.
- DeepSWE 1.173.7%
Software engineering tasks run in the same simple agent setup for every model.
10th of 27. The maker's own number, from Google, via LLM Stats.
- SimpleBench82.4%
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.
8th of 68. Independent result from Epoch AI.
- SimpleQA Verified69.7%
Short factual questions with one checked answer, asked without web search.
10th of 78. Independent result from Epoch AI.
8 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.