| Benchmark | Gemini 3.5 Flash | Gemini 3 Pro |
|---|---|---|
| AA Agentic Index | 70.4 | 52 |
| AA Intelligence | 52 | 40.6 |
| AA-LCR | 81 | 73 |
| AA-Omniscience | 21.2 | 15.3 |
| aime_2026 | 95 | 91.7 |
| apexAgents | 47.1 | 18.4 |
| arc_agi_1 | 92.5 | 75 |
| ARC-AGI-2 | 72.1 | 54 |
| arena_elo | 1477 | 1485 |
| arena_text_factuality | 1466 | 1481 |
| Artificial Analysis Coding Index | 70.1 | 46.5 |
| charxiv_rq | 84.2 | 81.4 |
| coding_arena_elo | 1490 | 1438 |
| critpt | 13.1 | 9.1 |
| frontiermath_tier_4 | 14.6 | 18.8 |
| gdpval | 57.8 | 34.2 |
| GDPval-AA (Elo) | 1656 | 1195 |
| GPQA Diamond | 92.2 | 91.9 |
| HLE | 42.7 | 45.8 |
| hmmt_feb_2026 | 95.5 | 86.4 |
| IFBench | 76.3 | 70.4 |
| livebench | 75 | 73.4 |
| matharena_visual_math_overall | 89.9 | 84.2 |
| MCP Atlas | 83.6 | 70.3 |
| MMMU-Pro | 84.3 | 81 |
| mrcr_v2_8needle_128k_average | 77.3 | 77 |
| mrcr_v2_8needle_1m_pointwise | 22.1 | 26.3 |
| OmniScience Accuracy | 51.4 | 55.8 |
| OmniScience Non-Hallucination | 38.2 | 10 |
| scicode | 53.1 | 56.1 |
| simplebench | 76.7 | 76.4 |
| simpleqa_verified | 68.4 | 72.9 |
| SWE-bench Pro | 55.1 | 43.3 |
| Terminal-Bench 2.0 | 76.2 | 54.2 |
| Terminal-Bench Hard | 46.2 | 56.9 |
| toolathlon | 56.5 | 36.4 |
| τ²-Bench Telecom (AA run) | 95.6 | 98 |
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.