GPT-5.5 (Non-reasoning): benchmark results
Provider: OpenAI. Released 2026-04-23. Access: API.
Aggregate Rating 1704 ± 12, rank #347 of 2170 rated models, from 119 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| PatternEval - Response-Pattern Failure Rate (Non-Thinking) | 5.04 | Bad-pattern trigger rate (%, lower is better; share of final | 100 |
| ROSE - Chinese Glyphs | 87.4 | Strict PASS (%) averaged over the five coupled tasks on the | 100 |
| ROSE - Emoji | 94.8 | Strict PASS (%) averaged over the five coupled tasks on the | 100 |
| ROSE - Exclusion Click | 92.5 | Strict PASS (%) on T5 exclusion clicking with count submissi | 100 |
| ROSE - Global Count | 93.8 | Strict PASS (%) on T1 global counting: the number of excepti | 100 |
| ROSE - Local Click | 93.6 | Strict PASS (%) on T3 local clicking: the exact coordinates | 100 |
| ROSE - Local Count | 97 | Strict PASS (%) on T2 local counting: the number of exceptio | 100 |
| ROSE - Pixel Art | 92.2 | Strict PASS (%) averaged over the five coupled tasks on the | 100 |
| ROSE - Visual-Region Click | 84.3 | Strict PASS (%) on T4 visual-region clicking: the exact coor | 100 |
| Review Error Injection (Zero-Shot) | 59.8 | Error detection recall (%; share of injected errors that the | 100 |
| RideWay | 70.5 | Efficiency Utility (success-gated, held-out custom14 mean ov | 100 |
| RuleWeaver - Same-Source - Rubric Score | 53.83 | Judge rubric score (0-100) | 100 |
Interactive version: theaggregate.ai/model?slug=gpt-5-5-non-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-05.