| Benchmark | Gemini 2.5 Flash | Qwen3.5 35B A3B |
|---|---|---|
| AA Agentic Index | 18.8 | 44.1 |
| AA Intelligence | 17.7 | 29.9 |
| AA-LCR | 65.7 | 68.3 |
| AA-Omniscience | -29.8 | -48.1 |
| Artificial Analysis Coding Index | 22.2 | 37 |
| critpt | 1.4 | 0.9 |
| gdpval | 11.9 | 20.3 |
| GPQA Diamond | 82.8 | 84.5 |
| HLE | 12.1 | 47.4 |
| hmmt_feb_2025 | 64.2 | 89 |
| IFBench | 50.3 | 72.5 |
| MMMU | 79.7 | 81.4 |
| MMMU-Pro | 69.1 | 75.1 |
| OmniScience Accuracy | 26.1 | 20.1 |
| OmniScience Non-Hallucination | 24.7 | 14.6 |
| scicode | 35.9 | 37.7 |
| SWE-bench Verified | 60.4 | 70 |
| Terminal-Bench Hard | 13.6 | 26.5 |
| vectara_answer_rate | 99 | 99.8 |
| vectara_avg_summary_length | 101.5 | 94.9 |
| vectara_factual_consistency | 92.2 | 89.5 |
| vectara_hallucination_rate ↓ | 7.8 | 10.5 |
| τ²-Bench Telecom (AA run) | 31.6 | 89.2 |
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.