Introducing GPT-5.5 (OpenAI) Apr 23, 2026
GPT-5.5 Pro
GPT-5.5 Pro ranks first on 3 of the 39 benchmarks with five or more models scored.
43
Benchmarks with a score
3
Ranked first, out of five or more models
12
In the top three
OpenAI Apr 2026
GPT-5.5 Pro
Closed weights
162.7
#14 of 256
Epoch Capabilities Index
- GPQA Diamond 93.9% #16 /229
- FrontierMath Tier 4 39.6% #2 /55
- ProtocolQA Open-Ended 36.7% #13 /27
Where it ranks
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.
Knowledge and language
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Humanity's Last Exam | 57.2% | #20 of 154 | Lab model card | |
| BullshitBench V2 | 36% | #69 of 129 | Official leaderboard |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| OTIS Mock AIME 2024-2025 | 100% | #2 of 212 | Epoch AI run | |
| FrontierMath | 52.4% | #1 of 70 | Epoch AI run | |
| FrontierMath Tier 4 | 39.6% | #2 of 55 | Epoch AI run | |
| IMO-ProofBench | 88.1% | #1 of 20 | Official leaderboard | |
| GPQA Diamond | 93.9% | #16 of 229 | Epoch AI run | |
| DTBench | 96% | #17 of 180 | Official leaderboard | |
| FrontierMath Tiers 1-3 v2 | 87.7% | #11 of 86 | Official leaderboard | |
| SimpleBench | 76.9% | #14 of 106 | Official leaderboard | |
| ARC-AGI | 96.5% | #13 of 95 | Official leaderboard | |
| ARC-AGI-2 | 84.6% | #15 of 98 | Official leaderboard | |
| FrontierMath Tier 4 v2 | 78% | #12 of 68 | Official leaderboard | |
| xbench-ScienceQA | 73% | #7 of 29 | Official leaderboard |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Chess Puzzles | 62% | #9 of 93 | Epoch AI run | |
| BrowseComp | 90.1% | #11 of 81 | Lab model card | |
| PinchBench 2.0 | 21.4% | #55 of 57 | Official leaderboard |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| prinzbench | 82.0 | #2 of 26 | Official leaderboard | |
| CADGenBench | 38.7% | #2 of 11 | Official leaderboard | |
| Investment Banking Modeling Tasks (OpenAI) | 88.6% | #1 of 5 | Lab model card | |
| GDPval | 82.3% | #4 of 10 | Lab model card |
Vision, audio and robotics
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MineBench | 1,903 | #12 of 75 | Official leaderboard | |
| VoxelBench | 1,963 | #13 of 56 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| ABC Bench Liquid-Handling Robot | 100% | #2 of 34 | Official leaderboard | |
| TroubleshootingBench | 49.6% | #2 of 20 | Lab model card | |
| Agentic BAIM-LLM Evaluation (ABLE) | 83.6% | #2 of 18 | Official leaderboard | |
| CVE-Bench | 93.1% | #2 of 14 | Lab model card | |
| Human Pathogen Capabilities Test v2 (HPCT v2) | 65.2% | #13 of 71 | Official leaderboard | |
| Tacit Knowledge and Troubleshooting | 81.7% | #6 of 30 | Lab model card | |
| DNA sequence design for transcription factor binding | 16.5% | #1 of 4 | Lab model card | |
| Multimodal Troubleshooting Virology | 55.3% | #8 of 27 | Lab model card | |
| BioTIER Refuse | 73.6% | #30 of 69 | Official leaderboard | |
| VCT-v2 | 45.3% | #12 of 25 | Official leaderboard | |
| ProtocolQA Open-Ended | 36.7% | #13 of 27 | Lab model card | |
| Biochemistry knowledge improvement | 39.3% | #1 of 2 | Lab model card | |
| BioTIER Permit | 99.2% | #43 of 69 | Official leaderboard | |
| ABC Bench Fragment Design | 87% | #17 of 27 | Official leaderboard | |
| Hard-negative protein binding prediction | 0% | #3 of 4 | Lab model card | |
| GeneBench | 33.2% | #2 of 2 | Lab model card |
Other
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Conceptual Reasoning Index | 66.2 | #18 of 151 | Official leaderboard | |
| LMCA | 53.9% | #20 of 147 | Official leaderboard | |
| ACCoRD | 75.8 | #28 of 152 | Official leaderboard | |
| StudentBench | 15.2 | #3 of 11 | Official leaderboard |
Its lab's numbers next to 2 leaderboards
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 |
|---|---|---|---|
| FrontierMath | 52.4% FrontierMath Tier 1-3 | 52.4% Epoch AI run, high | Same |
| FrontierMath Tier 4 | 39.6% FrontierMath Tier 4 | 39.6% Epoch AI run, xhigh | Same |
OpenAI's models over time
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)
Its launch post
The benchmarks the lab chose to headline, with the scores it gave.
Its system card
Every result in the longer documents the lab published with it. 8 of them aren't in the launch post.
GPT-5.5 System Card Apr 23, 2026
Bio and chem risk
Cyber risk
- CVE-Bench 93.1