GPT-5.4 (Non-reasoning): benchmark results
GPT-5.4 evaluated with reasoning disabled. Provider: OpenAI. Released 2026-03-06. Access: API.
Unified ELO 1628 ± 1, rank #326 of 1761 rated models, from 34 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| UGI - Natural Intelligence | 57.24 | NatInt Score | 92.8 |
| UGI - Writing | 57.99 | Writing Score | 92 |
| CritPt | 60 | Accuracy (self-reported) | 91.9 |
| AA Omniscience - Software Engineering (SWE) | 65.3 | Accuracy (%) | 87.5 |
| AA Terminal-Bench Hard | 37.88 | Accuracy (%) | 84.9 |
| UGI Leaderboard | 47.01 | UGI Score | 84 |
| AA Omniscience - Health | 34.7 | Accuracy (%) | 80.1 |
| AA-Omniscience Accuracy | 37.38 | Accuracy (%) | 79.9 |
| AA Omniscience - Law | 28.8 | Accuracy (%) | 79.7 |
| AA Omniscience - Business | 27.8 | Accuracy (%) | 78.1 |
| AA Omniscience - Humanities & Social Sciences | 31.7 | Accuracy (%) | 77.5 |
| Buyout Game (Lechmazur) | 1654 | Bradley-Terry Rating | 77.1 |
Interactive version: theaggregate.ai/model?slug=gpt-5-4-non-reasoning · How It Works · Data refreshed daily, snapshot 2026-09-05.