GPT-5.4 Pro ranks first on FrontierMath.
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 | |
|---|---|---|---|---|
| MultiNRC | 62.3% | #4 of 38 | Official leaderboard | |
| Humanity's Last Exam | 44.3% | #6 of 47 | Official leaderboard | |
| SkateBench v2 | 78.5% | #3 of 21 | Official leaderboard | |
| MultiChallenge | 69.2% | #9 of 42 | Official leaderboard | |
| Vectara Hallucination Leaderboard (2025 dataset) | 8.30% | #42 of 104 | Official leaderboard | |
| SimpleQA Verified | 46.3% | #41 of 94 | Epoch AI run |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| FrontierMath | 50% | #1 of 55 | Official leaderboard | |
| GPQA Diamond | 94.6% | #8 of 229 | Epoch AI run | |
| FrontierMath Tier 4 | 37.5% | #3 of 55 | Epoch AI run | |
| MathArena (Apex, IMO, USAMO) | 69.8% | #3 of 47 | Official leaderboard | |
| ARC-AGI-2 | 83.3% | #19 of 98 | Official leaderboard | |
| FrontierMath Tiers 1-3 v2 | 82.5% | #17 of 86 | Official leaderboard | |
| ARC-AGI | 94.5% | #21 of 95 | Official leaderboard | |
| HorizonMath | 7.10% | #2 of 8 | Official leaderboard | |
| FrontierMath Tier 4 v2 | 58.5% | #18 of 68 | Official leaderboard | |
| xbench-ScienceQA | 72.2% | #11 of 29 | Official leaderboard | |
| FrontierScience-Research | 36.7% | #2 of 5 | Lab model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Chess Puzzles | 59% | #11 of 93 | Epoch AI run | |
| BrowseComp | 89.3% | #12 of 81 | Lab model card | |
| PinchBench | 12% | #47 of 48 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| prinzbench | 79.0 | #3 of 26 | Official leaderboard | |
| TaxCalcBench | 62.8% | #2 of 17 | Official leaderboard | |
| Vals Finance Agent | 61.5% | #3 of 10 | Lab model card | |
| GDPval | 82% | #5 of 10 | Lab model card | |
| Investment Banking Modeling Tasks (OpenAI) | 83.6% | #3 of 5 | Lab model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| FORTRESS | 14.8% | #15 of 73 | Official leaderboard | |
| Human Pathogen Capabilities Test v2 (HPCT v2) | 63.9% | #20 of 71 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| TutorBench | 56.6% | #2 of 20 | 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 160.0 to 165.9. 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.
| Benchmark | Lab says | Board says | Gap |
|---|---|---|---|
| Humanity's Last Exam | 58.7% Best of Humanity's Last Exam no tools 42.7 / with tools 58.7 | 44.3% Leaderboard | +14 pts |
| FrontierMath Tier 4 | 38% FrontierMath Tier 4 | 37.5% Epoch AI run, web | +0.5 pts |
| GPQA Diamond | 94.4% GPQA Diamond | 94.6% Epoch AI run, xhigh | −0.2 pts |
| ARC-AGI-2 | 83.3% ARC-AGI-2 (Verified) | 83.3% Leaderboard, xhigh | Same |
| ARC-AGI | 94.5% ARC-AGI-1 (Verified) | 94.5% Leaderboard, xhigh | Same |
| FrontierMath | 50% FrontierMath Tier 1-3 | 50% Leaderboard, xhigh; task version 1.1.5, Epoch | Same |
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.