| Benchmark | Gemini 2.5 Flash | GPT-4o mini |
|---|---|---|
| AA Agentic Index | 18.8 | 1 |
| AA Intelligence | 17.7 | 7 |
| aider_polyglot | 61.9 | 3.6 |
| AIR-Bench 2024 | 68.7 | 56.3 |
| anthropic_red_team | 98.8 | 98.3 |
| ARC-AGI-2 | 2.5 | 0 |
| Artificial Analysis Coding Index | 22.2 | 11.4 |
| bbq | 97.7 | 88.2 |
| gdpval | 11.9 | 0 |
| GPQA Diamond | 82.8 | 42.6 |
| harmbench | 62.6 | 84.9 |
| HLE | 12.1 | 4.2 |
| IFBench | 50.3 | 31 |
| LiveCodeBench | 76.2 | 35.5 |
| livecodebench_easy | 99.1 | 81.7 |
| livecodebench_hard | 48.3 | 4.3 |
| livecodebench_medium | 82.5 | 25.2 |
| MMMU | 79.7 | 59.4 |
| MMMU-Pro | 69.1 | 41.5 |
| scicode | 35.9 | 22.9 |
| simple_safety_tests | 98 | 97.8 |
| simplebench | 41.2 | 10.7 |
| simpleqa | 26.9 | 9.9 |
| SWE-bench Verified | 60.4 | 8.7 |
| xstest | 98.8 | 96 |
Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing.