Gemini 3.1 Pro
#34 of 274 scored.
GoogleFeb 19, 2026Closed
Capabilities Index
Range 152.4–157.3
154.8
Context
Tokens
—
Input
Per 1M tokens
—
Output
Per 1M tokens
—
Benchmarks
How it scores.
- GPQA Diamond92.6%
7th of 186
- SWE-bench Verified75.6%
12th of 32
- FrontierMath59.7%
34th of 81
- ARC-AGI-277.1%
14th of 79
- Humanity's Last Exam43.7%
3rd of 40
- SimpleQA Verified73.5%
3rd of 77
- Mock AIME95.6%
31st of 176
- Terminal-Bench80.2%
3rd of 35
Everything else Epoch has for it
- ARC-AGI98.0%
- DTBench95.1%
- METR Time Horizons77.0%
- SimpleBench75.5%
- WeirdML72.1%
- LMCA63.3%
- Balrog57.0%
- Chess Puzzles52.6%
- DeepResearch Bench47.8%
- APEX-Agents35.3%
- Mystery Game Puzzles27.3%
- FrontierMath-Tier-4-v2-Private26.8%
- ExploitBench26.1%
- ProofBench26.0%
- GSO-Bench22.6%
- PostTrainBench22.0%
- CL-bench20.8%
- CL-bench Life16.9%
- EBR-bench14.3%
- DeepSWE11.7%
- MirrorCode8.9%
- Furniture Assembly0.0%
On the index
Around it.
Sources