| Benchmark | Claude 4 Opus | Gemini 3 Flash Preview |
|---|---|---|
| AA Intelligence | 31.7 | 38.7 |
| AA-LCR | 40 | 73 |
| AIME 2025 | 75.5 | 99.7 |
| arc_agi_1 | 27 | 84.7 |
| ARC-AGI-2 | 8.6 | 33.6 |
| arena_vision | 1207 | 1260 |
| Artificial Analysis Coding Index | 34 | 42.6 |
| frontiermath_tier_4 | 4.2 | 4.2 |
| GPQA Diamond | 79.6 | 90.4 |
| harmbench | 91.7 | 70 |
| HLE | 12.3 | 43.5 |
| hmmt_feb_2025 | 60 | 97.5 |
| IFBench | 53.7 | 78 |
| livebench | 74.6 | 72.4 |
| mmmlu | 88.8 | 91.8 |
| scicode | 40.9 | 50.6 |
| simplebench | 58.8 | 61.1 |
| simpleqa | 22.8 | 68.7 |
| swe_bench_bash | 67.6 | 75.8 |
| SWE-bench Verified | 79.4 | 78 |
| Terminal-Bench Hard | 31.1 | 38.6 |
| vectara_answer_rate | 91 | 99.8 |
| vectara_avg_summary_length | 123.2 | 90.2 |
| vectara_factual_consistency | 88 | 86.5 |
| vectara_hallucination_rate ↓ | 12 | 13.5 |
| τ²-Bench Telecom (AA run) | 73.4 | 80.4 |
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.