GPT-5.5 (xHigh): benchmark results
GPT-5.5 evaluated at the xhigh reasoning-effort setting. Provider: OpenAI. Released 2026-04-23. Access: API.
Aggregate Rating 1863 ± 5, rank #57 of 2168 rated models, from 459 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AGORA - Law | 62.86 | Accuracy (%) on the 35 law questions (285 documents), archiv | 100 |
| AGORA - Technology | 61.82 | Accuracy (%) on the 55 technology, software and manufacturin | 100 |
| AdvancedMathBench - ProverBench Qualifying Exam | 48.9 | Proof acceptance rate (%) on the 45 doctoral qualifying-exam | 100 |
| AdvancedMathBench - ProverBench Undergraduate | 64.5 | Proof acceptance rate (%) on the 200 undergraduate problems | 100 |
| AlmanBench - Acceptance | 95.43 | Acceptance Rate (%) | 100 |
| BackendForge | 28.57 | Task success rate (%; final oracle of 7,890 items including | 100 |
| Business Utility Eval (Codex) | 0.23 | Business Utility (0-1) | 100 |
| ClawProBench | 67.9 | Final Score (imported) | 100 |
| CoDA-Bench | 60.5 | Execution Accuracy (%) | 100 |
| GDPevo - self | 55.8 | Mean accuracy (%), self, 24 task groups, three repetitions | 100 |
| GRIPS | 96.4 | Accuracy (%) | 100 |
| GRIPS-hard | 87.3 | Accuracy (%) | 100 |
Interactive version: theaggregate.ai/model?slug=gpt-5-5-xhigh · How It Works · Data refreshed daily, snapshot 2026-10-20.