| Benchmark | Gemini 3 Pro | GLM-5.2 Full Open Source |
|---|---|---|
| AA Agentic Index | 52 | 45.7 |
| AA Intelligence | 40.6 | 52.6 |
| AA-LCR | 73 | 76.7 |
| AA-Omniscience | 15.3 | 4.4 |
| aime_2026 | 91.7 | 99.2 |
| apexAgents | 18.4 | 33.7 |
| arc_agi_1 | 75 | 77 |
| ARC-AGI-2 | 54 | 22.8 |
| arena_elo | 1485 | 1471 |
| arena_text_factuality | 1481 | 1460 |
| Artificial Analysis Coding Index | 46.5 | 68.8 |
| coding_arena_elo | 1438 | 1582 |
| critpt | 9.1 | 20.9 |
| gdpval | 34.2 | 50.3 |
| GPQA Diamond | 91.9 | 91.2 |
| HLE | 45.8 | 54.7 |
| HLE (with tools) | 45.8 | 54.7 |
| hmmt_feb_2026 | 86.4 | 92.5 |
| hmmt_nov_2025 | 93.3 | 94.4 |
| IFBench | 70.4 | 73.3 |
| imo_answer_bench | 83.3 | 91 |
| MCP Atlas | 70.3 | 82.6 |
| OmniScience Accuracy | 55.8 | 24.3 |
| OmniScience Non-Hallucination | 10 | 73.7 |
| scicode | 56.1 | 50.5 |
| simplebench | 76.4 | 58.8 |
| simpleqa_verified | 72.9 | 38.1 |
| SWE-bench Pro | 43.3 | 62.1 |
| Terminal-Bench Hard | 56.9 | 50.8 |
| tool_decathlon | 36.4 | 48.2 |
| toolathlon | 36.4 | 48.2 |
| τ²-Bench Telecom (AA run) | 98 | 99.1 |
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.