| Benchmark | Claude Opus 4.5 | Gemini 3.5 Flash |
|---|---|---|
| AA Agentic Index | 59.6 | 70.4 |
| AA Intelligence | 41.9 | 52 |
| AA-LCR | 76 | 81 |
| AA-Omniscience | 14 | 21.2 |
| aime_2026 | 95.1 | 95 |
| arc_agi_1 | 82 | 92.5 |
| ARC-AGI-2 | 37.6 | 72.1 |
| arena_elo | 1473 | 1477 |
| arena_text_factuality | 1458 | 1466 |
| Artificial Analysis Coding Index | 47.8 | 70.1 |
| charxiv_rq | 68.5 | 84.2 |
| coding_arena_elo | 1495 | 1490 |
| critpt | 4.6 | 13.1 |
| finance_agent | 55.2 | 57.9 |
| frontiermath_tier_4 | 4.2 | 14.6 |
| gdpval | 47.3 | 57.8 |
| GPQA Diamond | 87 | 92.2 |
| HLE | 30.8 | 42.7 |
| hmmt_feb_2026 | 85.3 | 95.5 |
| IFBench | 58 | 76.3 |
| livebench | 76 | 75 |
| livebench_agentic_coding | 39.7 | 49 |
| livebench_coding | 79.7 | 78.2 |
| livebench_data_analysis | 74.4 | 64.9 |
| livebench_instruction_following | 62.5 | 75.6 |
| livebench_language | 81.3 | 84.6 |
| livebench_math | 90.4 | 88.2 |
| livebench_reasoning | 80.1 | 82 |
| MCP Atlas | 69.8 | 83.6 |
| MMMU-Pro | 74 | 84.3 |
| OmniScience Accuracy | 46.6 | 51.4 |
| OmniScience Non-Hallucination | 39 | 38.2 |
| OSWorld-Verified | 66.3 | 78.4 |
| scicode | 50 | 53.1 |
| simplebench | 62 | 76.7 |
| simpleqa_verified | 41.8 | 68.4 |
| SWE-bench Pro | 57.1 | 55.1 |
| Terminal-Bench 2.0 | 59.3 | 76.2 |
| Terminal-Bench Hard | 47 | 46.2 |
| toolathlon | 43.5 | 56.5 |
| τ²-Bench Telecom (AA run) | 89.5 | 95.6 |
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.