GPT-5 ChatGPT (2025-08-07): benchmark results
Provider: OpenAI. Released 2025-08-07. Access: API.
Aggregate Rating 1693 ± 36, rank #383 of 10916 rated models, from 16 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| ClinSQL - Disease Diagnosis and Outcomes | 80.22 | Execution Score (%) | 100 |
| ClinSQL - Diagnostic Procedures | 75.38 | Execution Score (%) | 90.5 |
| M-JudgeBench - Logical Fallacy Detection | 97.9 | pairwise accuracy (%) on the 238 pairs where the rejected ch | 89.5 |
| M-JudgeBench - Short CoT Same-Model Pairs | 61.04 | pairwise accuracy (%) on the 326 pairs where a correct and a | 89.5 |
| ClinSQL | 68.42 | Execution Score (%) | 85.7 |
| ClinSQL - Laboratory Results Analysis | 57.89 | Execution Score (%) | 85.7 |
| ClinSQL - Medication Management | 71.07 | Execution Score (%) | 85.7 |
| ClinSQL - Patient Demographics and Admissions | 76.63 | Execution Score (%) | 85.7 |
| ClinSQL - SQL Score | 45.93 | Rubric SQL Score (%) | 85.7 |
| M-JudgeBench | 53.85 | Overall pairwise accuracy (%), pooled over all ten subtasks, | 84.2 |
| M-JudgeBench - Short CoT Cross-Model Pairs | 72.73 | pairwise accuracy (%) on the 297 pairs where a correct and a | 84.2 |
| M-JudgeBench - Long CoT Cross-Model Pairs | 61.14 | pairwise accuracy (%) on the 404 pairs where a correct and a | 78.9 |
Interactive version: theaggregate.ai/model?slug=gpt-5-chatgpt-2025-08-07 · How It Works · Data refreshed daily, snapshot 2026-11-08.