GPT-5.5 (High): benchmark results
GPT-5.5 evaluated at the high reasoning-effort setting. Provider: OpenAI. Released 2026-04-23. Access: API.
Aggregate Rating 1840 ± 7, rank #81 of 2168 rated models, from 341 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| BlueFin | 49.6 | Criteria-weighted rubric score (%; points met over maximum p | 100 |
| BlueFin - Interrogation | 50 | Criteria-weighted rubric score (%; points met over maximum p | 100 |
| Buyout Game (Lechmazur) | 1975.4 | Bradley-Terry Rating | 100 |
| CLBench Life | 22.2 | Solving Rate (%) | 100 |
| CLBench Life - Behavioral Records & Activity Trails | 25.4 | Solving Rate (%) | 100 |
| CLBench Life - Communication & Social Interactions | 23.5 | Solving Rate (%) | 100 |
| CLBench Life - Fragmented Information & Revisions | 17.8 | Solving Rate (%) | 100 |
| CoDyControlBench | 94.8 | Design success rate (%; Fixed-3 protocol: the model always r | 100 |
| CoDyControlBench (Success-Stop) | 91.7 | Design success rate (%; Success-Stop protocol: iteration sto | 100 |
| CoDyControlBench - PID-Family Controllers | 94.7 | Design success rate (%; PID family configurations under the | 100 |
| CoDyControlBench - SMC-Family Controllers | 94.9 | Design success rate (%; sliding-mode-control family configur | 100 |
| DPrivBench - Algorithm Level | 82.9 | F1 score (0-1, times 100) of the yes verdict on the DPrivBen | 100 |
Interactive version: theaggregate.ai/model?slug=gpt-5-5-high · How It Works · Data refreshed daily, snapshot 2026-10-20.