| Benchmark | Gemini 3.5 Flash | GPT-5.1 |
|---|---|---|
| AA Agentic Index | 70.4 | 32.2 |
| AA Intelligence | 52 | 37.5 |
| AA-LCR | 81 | 76.7 |
| AA-Omniscience | 21.2 | 5.4 |
| arc_agi_1 | 92.5 | 72.8 |
| ARC-AGI-2 | 72.1 | 17.6 |
| arena_elo | 1477 | 1439 |
| arena_text_factuality | 1466 | 1445 |
| Artificial Analysis Coding Index | 70.1 | 49.4 |
| coding_arena_elo | 1490 | 1421 |
| critpt | 13.1 | 4.9 |
| frontiermath_tier_4 | 14.6 | 12.5 |
| gdpval | 57.8 | 25 |
| GPQA Diamond | 92.2 | 88.1 |
| HLE | 42.7 | 28.5 |
| IFBench | 76.3 | 72.9 |
| matharena_visual_math_overall | 89.9 | 76.9 |
| MCP Atlas | 83.6 | 50.1 |
| MMMU-Pro | 84.3 | 75.5 |
| OmniScience Accuracy | 51.4 | 37.7 |
| OmniScience Non-Hallucination | 38.2 | 48.1 |
| scicode | 53.1 | 43.3 |
| simplebench | 76.7 | 53.2 |
| simpleqa_verified | 68.4 | 48.9 |
| TauBench V3 - Banking | 32.2 | 15.9 |
| Terminal-Bench 2.0 | 76.2 | 47.6 |
| Terminal-Bench 2.1 | 78.7 | 52.4 |
| Terminal-Bench Hard | 46.2 | 45.5 |
| τ²-Bench Telecom (AA run) | 95.6 | 81.9 |
| τ³-Bench | 25.4 | 14 |
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.