GPT-5.5 Instant (OpenAI) May 5, 2026
GPT-5.5 Instant
GPT-5.5 Instant places in the top three on Capture the Flag (Professional).
26
Benchmarks with a score
0
Ranked first, out of five or more models
1
In the top three
OpenAI May 2026
GPT-5.5 Instant
Closed weights
143.8
#146 of 404
Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.
- MMMU-Pro 76% #35 /90
- GPQA Diamond 82.5% #90 /229
- HealthBench Professional 38.4% #25 /26
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 | |
|---|---|---|---|---|
| LMArena (Elo) | 1,473 | #37 of 335 | Official leaderboard | |
| Document Arena | 1,460 | #22 of 39 | Official leaderboard | |
| OpenAI user-flagged hallucination eval | 46.1% | #4 of 4 | Lab model card | |
| Factuality Heavy prompts | 17.9% | #3 of 3 | Lab model card | |
| High Stakes factuality prompts | 19.9% | #3 of 3 | Lab model card |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| GPQA Diamond | 82.5% | #90 of 229 | Epoch AI run | |
| OTIS Mock AIME 2024-2025 | 68.1% | #114 of 212 | Epoch AI run | |
| AIME 2025 | 81.2% | #87 of 143 | Lab model card | |
| FrontierMath Tiers 1-3 v2 | 26.3% | #69 of 86 | Official leaderboard | |
| FrontierMath Tier 4 v2 | 2.44% | #65 of 68 | Official leaderboard |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| HealthBench Consensus | 94.7% | #13 of 17 | Lab model card | |
| HealthBench | 51.4% | #29 of 31 | Lab model card | |
| HealthBench Hard | 22.9% | #20 of 21 | Lab model card | |
| HealthBench Professional | 38.4% | #25 of 26 | Lab model card |
Vision, audio and robotics
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Vision Arena | 1,277 | #26 of 136 | Official leaderboard | |
| CharXiv | 81.6% | #25 of 75 | Lab model card | |
| MMMU-Pro | 76% | #35 of 90 | Lab model card |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Capture the Flag (Professional) | 94.1% | #2 of 21 | Lab model card | |
| CVE-Bench | 90.2% | #4 of 14 | Lab model card | |
| Cyber Range | 76.9% | #4 of 10 | Lab model card | |
| ProtocolQA Open-Ended | 36.7% | #14 of 27 | Lab model card | |
| Multimodal Troubleshooting Virology | 51.1% | #16 of 27 | Lab model card | |
| TroubleshootingBench | 33.7% | #13 of 20 | Lab model card | |
| Tacit Knowledge and Troubleshooting | 68.9% | #20 of 30 | Lab model card | |
| Agent JSK | 99.2% | #5 of 7 | Lab model card | |
| First-Person Fairness evaluation | 1.01% | #9 of 10 | Lab model card |
Its lab's numbers put it at 143.8, number 146 of 404
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 139.9 to 147.7. How we compute it.
- HealthBench Consensus 94.7
- GPQA Diamond 85.6
- CharXiv 81.6
- AIME 2025 81.2
- MMMU-Pro 76
- HealthBench 51.4
- HealthBench Professional 38.4
- HealthBench Hard 22.9
Its lab's number next to a leaderboard
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 |
|---|---|---|---|
| GPQA Diamond | 85.6% GPQA (PhD-Level Science) | 82.5% Epoch AI run | +3.1 pts |
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. 22 of them aren't in the launch post.
GPT-5.5 Instant System Card May 4, 2026
Health
Hallucination
Safety
Robustness
Bias
Bio and chem risk
Cyber risk
- Capture the Flag (Professional) 94.1
- CVE-Bench 90.2
- Cyber Range 76.9