| Benchmark | Gemini 3 Flash Preview | GPT-4o |
|---|---|---|
| AA Agentic Index | 49.7 | 9.7 |
| AA Intelligence | 38.7 | 11.2 |
| AA-LCR | 73 | 55 |
| AA-Omniscience | 10.1 | -10.5 |
| AIME 2025 | 99.7 | 11.7 |
| arc_agi_1 | 84.7 | 4.5 |
| ARC-AGI-2 | 33.6 | 0 |
| arena_vision | 1260 | 1162 |
| Artificial Analysis Coding Index | 42.6 | 24.2 |
| critpt | 8.6 | 0 |
| gaia_level1 | 82.8 | 39.8 |
| gaia_level2 | 66 | 27 |
| gaia_level3 | 49 | 14.3 |
| gaia_overall | 68.4 | 28.9 |
| gdpval | 35.2 | 0 |
| GPQA Diamond | 90.4 | 70.1 |
| harmbench | 70 | 82.9 |
| HLE | 43.5 | 5.3 |
| IFBench | 78 | 36 |
| mmmlu | 91.8 | 81.4 |
| MMMU-Pro | 81.2 | 59.9 |
| OmniScience Accuracy | 53.4 | 23.7 |
| OmniScience Non-Hallucination | 7.6 | 62.1 |
| OSWorld-Verified | 65.1 | 5 |
| scicode | 50.6 | 33.4 |
| simplebench | 61.1 | 17.8 |
| simpleqa | 68.7 | 39.4 |
| swe_bench_bash | 75.8 | 21.6 |
| SWE-bench Verified | 78 | 38.8 |
| Terminal-Bench Hard | 38.6 | 8.3 |
| vectara_answer_rate | 99.8 | 93.8 |
| vectara_avg_summary_length | 90.2 | 86.6 |
| vectara_factual_consistency | 86.5 | 90.4 |
| vectara_hallucination_rate ↓ | 13.5 | 9.6 |
| VideoMMMU | 86.9 | 61.2 |
| τ²-Bench Telecom (AA run) | 80.4 | 28.9 |
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.