GPT-5.2 Instant: benchmark results
Provider: OpenAI. Released 2025-12-11. Access: API.
Aggregate Rating 1694 ± 11, rank #207 of 1635 rated models, from 160 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| ALPBench - Cell Phones (Attribute F1) | 46.61 | Attribute-level F1 (%): the F1 of each target attribute comp | 100 |
| EHRBench - Diagnosis | 72.02 | Accuracy (%) on the diagnosis-decision questions (complete a | 100 |
| EHRBench - MIMIC-III | 71.5 | Accuracy (%) on the questions built from MIMIC-III records, | 100 |
| EHRBench - MIMIC-IV | 71.06 | Accuracy (%) on the questions built from MIMIC-IV records, E | 100 |
| EHRBench - PROMOTE | 70.7 | Accuracy (%) on the questions built from PROMOTE records, EH | 100 |
| EHRBench - Prognosis | 60.59 | Accuracy (%) on the prognosis-decision questions (predict a | 100 |
| EgoPoint-Bench - Real-World Adversarial Resilience | 71.27 | Accuracy (%) on the Adversarial Resilience questions among t | 100 |
| EgoPoint-Bench - Real-World Spatial Context | 81.62 | Accuracy (%) on the Spatial Context questions among the 1,16 | 100 |
| MHGraphBench - Relation Prediction | 58.63 | Accuracy (%) on the 1,634 four-way drug-disease relation pre | 100 |
| MHGraphBench - Two-hop Selection | 65.58 | Accuracy (%) on the 1,200 two-hop selection items, letter-on | 100 |
| MHGraphBench - Two-hop Selection with Evidence | 67.58 | Accuracy (%) on the 1,200 two-hop selection items with sanit | 100 |
| EHRBench - Treatment | 80.13 | Accuracy (%) on the treatment-decision questions, EHR-ground | 96.7 |
Interactive version: theaggregate.ai/model?slug=gpt-5-2-instant · How It Works · Data refreshed daily, snapshot 2026-10-15.