| Benchmark | Gemini 3.6 Flash | GLM-5.2 Full Open Source |
|---|---|---|
| AA Agentic Index | 40.5 | 45.7 |
| AA Intelligence | 52 | 52.6 |
| AA-LCR | 79 | 76.7 |
| AA-Omniscience | 22.1 | 4.4 |
| aime_2026 | 96.7 | 99.2 |
| arc_agi_1 | 34.5 | 77 |
| ARC-AGI-2 | 2.6 | 22.8 |
| arena_elo | 1481 | 1471 |
| arena_text_factuality | 1470 | 1460 |
| Artificial Analysis Coding Index | 69.2 | 68.8 |
| coding_arena_elo | 1538 | 1582 |
| critpt | 10.6 | 20.9 |
| DeepSWE | 49 | 46.2 |
| DeepSWE 1.1 | 49 | 44 |
| gdpval | 46.1 | 50.3 |
| GDPval-AA v2 | 1421 | 1514 |
| GPQA Diamond | 92.8 | 91.2 |
| HLE | 40.8 | 54.7 |
| hmmt_feb_2026 | 89.4 | 92.5 |
| livebench_agentic_coding | 43.4 | 51.8 |
| livebench_coding | 77.9 | 79.7 |
| livebench_data_analysis | 63 | 73.7 |
| livebench_instruction_following | 75.4 | 62.3 |
| livebench_language | 83.9 | 76.2 |
| livebench_math | 86.4 | 89.8 |
| livebench_reasoning | 85.2 | 78.6 |
| OmniScience Accuracy | 50 | 24.3 |
| OmniScience Non-Hallucination | 44.4 | 73.7 |
| scicode | 52.7 | 50.5 |
| simpleqa_verified | 68.7 | 38.1 |
| SWE-bench Pro | 58.7 | 62.1 |
| TauBench V3 - Banking | 29.9 | 34.6 |
| Terminal-Bench 2.1 | 78 | 82.7 |
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.