| Benchmark | Gemini 3 Flash Preview | GPT-5.1 |
|---|---|---|
| AA Agentic Index | 49.7 | 32.2 |
| AA Intelligence | 38.7 | 37.5 |
| AA-LCR | 73 | 76.7 |
| AA-Omniscience | 10.1 | 5.4 |
| AIME 2025 | 99.7 | 94.2 |
| arc_agi_1 | 84.7 | 72.8 |
| ARC-AGI-2 | 33.6 | 17.6 |
| arena_elo | 1473 | 1439 |
| arena_text_factuality | 1452 | 1445 |
| arena_vision | 1260 | 1250 |
| Artificial Analysis Coding Index | 42.6 | 49.4 |
| coding_arena_elo | 1387 | 1421 |
| critpt | 8.6 | 4.9 |
| frontiermath_tier_4 | 4.2 | 12.5 |
| gdpval | 35.2 | 25 |
| GPQA Diamond | 90.4 | 88.1 |
| HLE | 43.5 | 28.5 |
| hmmt_feb_2025 | 97.5 | 96.3 |
| hmmt_nov_2025 | 93.3 | 91.7 |
| IFBench | 78 | 72.9 |
| matharena_visual_math_overall | 85.8 | 76.9 |
| MCP Atlas | 62 | 50.1 |
| mmmlu | 91.8 | 91 |
| MMMU-Pro | 81.2 | 75.5 |
| OmniScience Accuracy | 53.4 | 37.7 |
| OmniScience Non-Hallucination | 7.6 | 48.1 |
| scicode | 50.6 | 43.3 |
| simplebench | 61.1 | 53.2 |
| simpleqa_verified | 67.4 | 48.9 |
| swe_bench_bash | 75.8 | 66 |
| SWE-bench Verified | 78 | 76.3 |
| TauBench V3 - Banking | 20.8 | 15.9 |
| Terminal-Bench 2.0 | 47.6 | 47.6 |
| Terminal-Bench 2.1 | 58 | 52.4 |
| Terminal-Bench Hard | 38.6 | 45.5 |
| vectara_answer_rate | 99.8 | 100 |
| vectara_avg_summary_length | 90.2 | 254.4 |
| vectara_factual_consistency | 86.5 | 89.1 |
| vectara_hallucination_rate ↓ | 13.5 | 12.1 |
| τ²-Bench Telecom (AA run) | 80.4 | 81.9 |
| τ³-Bench | 17.5 | 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.