GPT-5 (Minimal): benchmark results
GPT-5 evaluated at the minimal reasoning-effort setting. Provider: OpenAI. Released 2025-08-07. Access: API.
Aggregate Rating 1599 ± 11, rank #718 of 2168 rated models, from 137 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Portuguese LLM - FaQuAD NLI | 90.49 | Macro F1 (%) | 100 |
| PriVE-Bench | 46.7 | Counterfactual accuracy (%; macro-averaged over the five dom | 100 |
| PriVE-Bench - Fashion Attributes | 44.3 | Counterfactual accuracy (%) in the fashion attributes domain | 100 |
| PriVE-Bench - Industry Attributes | 85.1 | Counterfactual accuracy (%) in the industry attributes domai | 100 |
| SQuID | 28.1 | Accuracy (%; answer inside the human-variance range or exact | 100 |
| Open Portuguese LLM - ASSIN2 RTE | 94.98 | Macro F1 (%) | 99.8 |
| Open Portuguese LLM - OAB Exams | 81.05 | Accuracy (%) | 99.4 |
| Open Portuguese LLM - BLUEX | 78.86 | Accuracy (%) | 98.2 |
| Open Portuguese LLM - TweetSentBR | 78.78 | F1 (%) | 97.6 |
| Open Portuguese LLM - ENEM | 84.32 | Accuracy (%) | 97.2 |
| Diplomacy: Steerability | 21.9 | Mean game-score difference: optimized aggressive prompt minu | 95.2 |
| Open Portuguese LLM - HateBR Offensive | 92.33 | F1 (%) | 90.2 |
Interactive version: theaggregate.ai/model?slug=gpt-5-minimal · How It Works · Data refreshed daily, snapshot 2026-10-20.