| Benchmark | Claude Opus 4.7 | Gemini 2.5 Flash |
|---|---|---|
| AA Agentic Index | 64.6 | 18.8 |
| AA Intelligence | 55 | 17.7 |
| AA-LCR | 75.3 | 65.7 |
| AA-Omniscience | 27.3 | -29.8 |
| AIME | 91 | 56.3 |
| arc_agi_1 | 92 | 33.3 |
| ARC-AGI-2 | 75.8 | 2.5 |
| arena_text_factuality | 1474 | 1437 |
| arena_vision | 1304 | 1214 |
| Artificial Analysis Coding Index | 73.6 | 22.2 |
| critpt | 12 | 1.4 |
| frontiermath_tier_4 | 22.9 | 4.2 |
| gdpval | 58.6 | 11.9 |
| GPQA Diamond | 94.2 | 82.8 |
| harmbench | 17.5 | 62.6 |
| HLE | 54.7 | 12.1 |
| IFBench | 58.6 | 50.3 |
| MMMU-Pro | 78.8 | 69.1 |
| OmniScience Accuracy | 48.9 | 26.1 |
| OmniScience Non-Hallucination | 57.7 | 24.7 |
| scicode | 54.5 | 35.9 |
| simplebench | 61.7 | 41.2 |
| SWE-bench Verified | 87.6 | 60.4 |
| Terminal-Bench Hard | 54.5 | 13.6 |
| vectara_answer_rate | 98 | 99 |
| vectara_avg_summary_length | 149.1 | 101.5 |
| vectara_factual_consistency | 88 | 92.2 |
| vectara_hallucination_rate ↓ | 12 | 7.8 |
| τ²-Bench Telecom (AA run) | 88.6 | 31.6 |
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.