| Benchmark | Claude 4 Opus | Gemma 4 31B |
|---|---|---|
| AA Intelligence | 31.7 | 29.7 |
| AA-LCR | 40 | 68.3 |
| arena_vision | 1207 | 1252 |
| Artificial Analysis Coding Index | 34 | 43.4 |
| GPQA Diamond | 79.6 | 85.7 |
| HLE | 12.3 | 26.5 |
| hmmt_feb_2025 | 60 | 88.7 |
| IFBench | 53.7 | 75.6 |
| LiveCodeBench v6 | 47.4 | 80 |
| mmlu_redux | 94.2 | 93.7 |
| MMLU-Pro | 86.6 | 85.2 |
| mmmlu | 88.8 | 88.4 |
| MMMU | 73.7 | 80.4 |
| scicode | 40.9 | 43.4 |
| supergpqa | 56.5 | 65.7 |
| SWE-bench Verified | 79.4 | 52 |
| Terminal-Bench Hard | 31.1 | 36.4 |
| vectara_answer_rate | 91 | 100 |
| vectara_avg_summary_length | 123.2 | 75.8 |
| vectara_factual_consistency | 88 | 92.6 |
| vectara_hallucination_rate ↓ | 12 | 7.4 |
| τ²-Bench | 57 | 76.9 |
| τ²-Bench Telecom (AA run) | 73.4 | 65.5 |
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.