| Benchmark | Gemini 3 Flash Preview | GLM-5.1 |
|---|---|---|
| AA Agentic Index | 49.7 | 66 |
| AA Intelligence | 38.7 | 41 |
| AA-LCR | 73 | 68 |
| AA-Omniscience | 10.1 | 0.8 |
| aime_2026 | 96.7 | 95.8 |
| arena_elo | 1473 | 1468 |
| arena_text_factuality | 1452 | 1458 |
| Artificial Analysis Coding Index | 42.6 | 55.8 |
| coding_arena_elo | 1387 | 1510 |
| critpt | 8.6 | 4.6 |
| Finance Agent v2 | 42.6 | 44.8 |
| frontiermath_tier_4 | 4.2 | 12.5 |
| gdpval | 35.2 | 49.5 |
| GDPval-AA (Elo) | 1204 | 1535 |
| GPQA Diamond | 90.4 | 86.8 |
| HLE | 43.5 | 52.3 |
| hmmt_feb_2026 | 89.4 | 89.4 |
| hmmt_nov_2025 | 93.3 | 94 |
| IFBench | 78 | 76.3 |
| livebench | 72.4 | 70.2 |
| MCP Atlas | 62 | 75.6 |
| OmniScience Accuracy | 53.4 | 25.2 |
| OmniScience Non-Hallucination | 7.6 | 70.1 |
| scicode | 50.6 | 43.8 |
| simplebench | 61.1 | 55.1 |
| simpleqa_verified | 67.4 | 38.1 |
| SWE-bench Multilingual | 72.7 | 73.3 |
| SWE-bench Pro | 49.6 | 58.4 |
| TauBench V3 - Banking | 20.8 | 13.6 |
| Terminal-Bench 2.0 | 47.6 | 69 |
| Terminal-Bench 2.1 | 58 | 63.5 |
| Terminal-Bench Hard | 38.6 | 43.2 |
| toolathlon | 49.4 | 40.7 |
| Vending-Bench 2 | 363500 | 563441 |
| τ²-Bench Telecom (AA run) | 80.4 | 97.7 |
| τ³-Bench | 17.5 | 70.6 |
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.