GPT-5 Chat: benchmark results
OpenAI's non-reasoning, ChatGPT-tuned GPT-5 snapshot exposed in the API as gpt-5-chat. Provider: OpenAI. Released 2025-08-07. Access: API.
Aggregate Rating 1682 ± 10, rank #241 of 1656 rated models, from 216 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AmongAIs | 750 | Score (50×Impostor + 10×Crewmate wins) | 100 |
| CVSBench - FOV Satellite-to-Ground | 54.1 | Accuracy (%) on FOV-subset satellite-to-ground questions (li | 100 |
| EHRBench - Treatment | 80.45 | Accuracy (%) on the treatment-decision questions, EHR-ground | 100 |
| HelpBench - Harassment | 88 | Harassment topic (interpersonal harassment and abuse such as | 100 |
| HelpBench - Platform Actions | 89 | Platform Actions topic (platform or third-party safety actio | 100 |
| HelpBench - Privacy Tools | 86 | Privacy Tools topic (privacy tools such as ad blockers, encr | 100 |
| HelpBench - Scams | 93 | Scams topic (scams, fraud, spam and suspicious artifacts) ru | 100 |
| PreScam - Psychological Technique Prediction (Limited) | 50.91 | PT HitRate (%): the share of gold psychological techniques o | 100 |
| StratMem-Bench - Memory Integration Quality | 4.56 | Memory integration quality MIQ (1-5, judge-rated) of respons | 100 |
| AgentDrive - Policy | 100 | Accuracy (%) | 99 |
| BenchTable - Utility | 83.6 | Weighted Score (%) | 98.8 |
| AgentDrive - Hybrid | 70 | Accuracy (%) | 98 |
Interactive version: theaggregate.ai/model?slug=gpt-5-chat · How It Works · Data refreshed daily, snapshot 2026-10-20.