| Benchmark | Claude Opus 4.8 | Gemini 2.5 Pro |
|---|---|---|
| AA Agentic Index | 49.4 | 7.2 |
| AA Intelligence | 57.3 | 27 |
| AA-LCR | 73 | 66 |
| AA-Omniscience | 28.8 | -14.3 |
| arc_agi_1 | 92.5 | 41 |
| ARC-AGI-2 | 72.1 | 4.9 |
| arena_elo | 1482 | 1446 |
| arena_text_factuality | 1464 | 1462 |
| arena_vision | 1280 | 1246 |
| Artificial Analysis Coding Index | 74.3 | 46.7 |
| browsecomp | 88.5 | 9.9 |
| coding_arena_elo | 1539 | 1224 |
| critpt | 20.9 | 2.6 |
| Fortress | 18.2 | 54.9 |
| frontiermath_tier_4 | 31.3 | 4.2 |
| gdpval | 54.2 | 8.5 |
| GPQA Diamond | 93.6 | 84.4 |
| HLE | 57.9 | 22.5 |
| HLE (with tools) | 57.9 | 28.4 |
| hmmt_nov_2025 | 96.5 | 80 |
| IFBench | 62.2 | 49 |
| matharena_visual_math_overall | 81.6 | 77.2 |
| mathvision | 86.7 | 73.3 |
| MMMU-Pro | 78.9 | 74.9 |
| OmniScience Accuracy | 48.8 | 39 |
| OmniScience Non-Hallucination | 60.7 | 12.6 |
| scicode | 53.5 | 43 |
| simplebench | 64.8 | 62.4 |
| simpleqa_verified | 39.5 | 56 |
| SWE-bench Verified | 88.6 | 67.2 |
| TauBench V3 - Banking | 34.2 | 9.7 |
| Terminal-Bench 2.1 | 85 | 28.5 |
| Terminal-Bench Hard | 58.3 | 26.5 |
| Video-MME | 86 | 84.8 |
| τ²-Bench Telecom (AA run) | 94.4 | 54.1 |
| τ³-Bench | 27.6 | 9.3 |
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.