GPT-5.2 (xHigh): benchmark results
GPT-5.2 evaluated at the xhigh reasoning-effort setting. Provider: OpenAI. Released 2025-12-11. Access: API.
Aggregate Rating 1795 ± 12, rank #148 of 2168 rated models, from 149 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| LongCoT - Chess | 7.4 | Accuracy (%) on the medium and hard chess problems of LongCo | 100 |
| LongCoT - Computer Science | 9.7 | Accuracy (%) on the medium and hard computer science problem | 100 |
| LongCoT - Logic | 12 | Accuracy (%) on the medium and hard logic problems of LongCo | 100 |
| LongCoT-mini - Chess | 36.7 | Accuracy (%) on the easy chess problems of LongCoT-mini (lon | 100 |
| LongCoT-mini - Computer Science | 44.4 | Accuracy (%) on the easy computer science problems of LongCo | 100 |
| LongCoT-mini - Logic | 53.6 | Accuracy (%) on the easy logic problems of LongCoT-mini (lon | 100 |
| MMErroR - Error Presence Detection - Data & Analytics | 65.88 | Error Presence Detection Accuracy (%) | 100 |
| MMErroR - Error Type Classification - Data & Analytics | 69.59 | Error Type Classification Accuracy (%) | 100 |
| MathArena - HMMT Feb 2025 | 100 | Accuracy (%) | 100 |
| MathArena - HMMT Nov 2025 | 99.17 | Accuracy (%) | 100 |
| MultihopSpatial - 2-Hop Ego-Centric | 49.7 | Multiple-choice accuracy (%) on the 2-hop ego-centric questi | 100 |
| Pencil Puzzle Bench - Firefly | 33.33 | Direct-ask Success Rate (%) | 100 |
Interactive version: theaggregate.ai/model?slug=gpt-5-2-xhigh · How It Works · Data refreshed daily, snapshot 2026-10-20.