| Benchmark | GPT-5.2 | o3 |
|---|---|---|
| AA Agentic Index | 60.2 | 36.1 |
| AA Intelligence | 43.3 | 31.1 |
| AA-LCR | 79.3 | 73.3 |
| AA-Omniscience | 0.3 | -15.3 |
| AIME 2025 | 100 | 89.2 |
| arc_agi_1 | 86.2 | 60.8 |
| ARC-AGI-2 | 52.9 | 6.5 |
| arena_vision | 1245 | 1217 |
| Artificial Analysis Coding Index | 48.7 | 38.4 |
| browsecomp | 65.8 | 49.7 |
| critpt | 11.6 | 1.1 |
| Fortress | 17.5 | 16 |
| frontiermath_tier_4 | 18.8 | 2.1 |
| gdpval | 48.3 | 12.8 |
| GPQA Diamond | 92.4 | 83.3 |
| HLE | 45.5 | 20.3 |
| hmmt_feb_2025 | 100 | 77.5 |
| IFBench | 75.4 | 71.4 |
| LiveCodeBench | 89 | 84.7 |
| longbench_v2 | 54.5 | 58.8 |
| MathVista | 82.8 | 86.8 |
| metr_hcast | 77 | 64.7 |
| metr_re_bench | 35 | 15 |
| metr_swaa | 99.6 | 99.8 |
| MMLU-Pro | 87 | 85 |
| MMMU-Pro | 79.5 | 76.4 |
| MultiNRC | 42.2 | 45.5 |
| OmniScience Accuracy | 44.3 | 38.6 |
| OmniScience Non-Hallucination | 38.4 | 12.9 |
| PropensityBench | 34.4 | 10.5 |
| scicode | 52.1 | 41 |
| simplebench | 45.8 | 53.1 |
| simpleqa_verified | 38.9 | 53 |
| swe_bench_bash | 72.8 | 58.4 |
| SWE-bench Verified | 80 | 69.1 |
| Tau2 retail | 82 | 80.2 |
| Terminal-Bench Hard | 62.2 | 37.1 |
| VideoMMMU | 85.9 | 83.3 |
| τ²-Bench Telecom (AA run) | 98.7 | 80.7 |
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.