GPT-5.2 Pro ranks first on LABBench2 LitQA3.
Closed weights
Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| SuperGPQA | 67.1% | #2 of 59 | Official leaderboard | |
| Humanity's Last Exam | 50% | #42 of 154 | Lab model card | |
| SkateBench v2 | 71.8% | #8 of 21 | Official leaderboard | |
| Extended NYT Connections | 79.3% | #44 of 103 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| GPQA Diamond | 93.2% | #12 of 302 | Lab model card | |
| AIME 2025 | 100% | #7 of 143 | Lab model card | |
| HMMT 2025 | 100% | #3 of 54 | Lab model card | |
| FrontierMath Tier 4 | 31.3% | #5 of 55 | Epoch AI run | |
| FrontierMath Tiers 1-3 v2 | 74% | #25 of 86 | Official leaderboard | |
| SGI-Bench | 31.1% | #5 of 15 | Official leaderboard | |
| IMProofBench | 36.9% | #3 of 9 | Official leaderboard | |
| ARC-AGI | 90.5% | #33 of 95 | Official leaderboard | |
| FrontierMath Tier 4 v2 | 46% | #26 of 68 | Official leaderboard | |
| LABBench2 | 58.3% | #2 of 5 | Official leaderboard | |
| ARC-AGI-2 | 54.2% | #42 of 98 | Official leaderboard | |
| SimpleBench | 57.4% | #49 of 106 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Chess Puzzles | 60% | #10 of 93 | Epoch AI run | |
| Pencil Puzzle Bench | 26.7% | #10 of 53 | Official leaderboard | |
| BrowseComp | 77.9% | #33 of 81 | Lab model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| TaxCalcBench | 41.2% | #6 of 17 | Official leaderboard | |
| PRBench | 45.4% | #25 of 48 | Official leaderboard | |
| GDPval | 74.1% | #6 of 10 | Lab model card | |
| Investment Banking Modeling Tasks (OpenAI) | 71.7% | #4 of 5 | Lab model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MineBench | 1,539 | #37 of 75 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LABBench2 LitQA3 | 85.1% | #1 of 5 | Official leaderboard | |
| World-Class Bio | 56.3% | #16 of 74 | Official leaderboard | |
| Human Pathogen Capabilities Test v2 (HPCT v2) | 64% | #19 of 71 | Official leaderboard | |
| LABBench2 FigQA2 | 34.7% | #2 of 5 | Official leaderboard | |
| LABBench2 SeqQA2 | 46.3% | #3 of 5 | Official leaderboard | |
| LABBench2 ProtocolQA2 | 50% | #4 of 5 | Official leaderboard |
The Self-Reported Index fits one score to the 8 capability results OpenAI reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 152.7 to 157.7. How we compute it.
Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.
OpenAI's models in color, against every other model on the Epoch Capabilities Index, by release date.
Source: Epoch Capabilities Index, Epoch AI (CC BY 4.0)
The benchmarks the lab chose to headline, with the scores it gave.