| Benchmark | Gemini 3 Flash Preview | Qwen3.6 Plus |
|---|---|---|
| AA Agentic Index | 49.7 | 29 |
| AA Intelligence | 38.7 | 40.5 |
| AA-LCR | 73 | 72.3 |
| AA-Omniscience | 10.1 | 2.6 |
| aime_2026 | 96.7 | 95.3 |
| arena_elo | 1473 | 1444 |
| arena_text_factuality | 1452 | 1438 |
| Artificial Analysis Coding Index | 42.6 | 54.5 |
| coding_arena_elo | 1387 | 1460 |
| critpt | 8.6 | 2.9 |
| Finance Agent v2 | 42.6 | 40.9 |
| frontiermath_tier_4 | 4.2 | 8.3 |
| gdpval | 35.2 | 32 |
| GPQA Diamond | 90.4 | 90.4 |
| HLE | 43.5 | 28.8 |
| hmmt_feb_2026 | 89.4 | 87.8 |
| hmmt_nov_2025 | 93.3 | 94.6 |
| IFBench | 78 | 75.2 |
| livebench | 72.4 | 70.8 |
| MCP Atlas | 62 | 74.1 |
| mmmlu | 91.8 | 89.5 |
| MMMU-Pro | 81.2 | 78.8 |
| OmniDocBench 1.5 | 12.1 | 91.2 |
| OmniScience Accuracy | 53.4 | 26.4 |
| OmniScience Non-Hallucination | 7.6 | 68 |
| OSWorld-Verified | 65.1 | 62.5 |
| scicode | 50.6 | 40.7 |
| screenspot_pro_no_tools | 69.1 | 68.2 |
| simpleqa_verified | 67.4 | 49.1 |
| SWE-bench Multilingual | 72.7 | 73.8 |
| SWE-bench Pro | 49.6 | 56.6 |
| SWE-bench Verified | 78 | 78.8 |
| TauBench V3 - Banking | 20.8 | 20.8 |
| Terminal-Bench 2.0 | 47.6 | 61.6 |
| Terminal-Bench 2.1 | 58 | 61.4 |
| Terminal-Bench Hard | 38.6 | 43.9 |
| toolathlon | 49.4 | 39.8 |
| VideoMMMU | 86.9 | 84 |
| τ²-Bench Telecom (AA run) | 80.4 | 97.7 |
| τ³-Bench | 17.5 | 70.7 |
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.