GPT-5.4 Mini (2026-03-17): benchmark results
Provider: OpenAI. Released 2026-03-17. Access: API.
Unified ELO 1721 ± 10, rank #155 of 1639 rated models, from 613 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AuthMem-Bench (Task Success Rate) | 56.6 | Task success rate (%; share of 350 authorized counterpart ca | 100 |
| EuroEval Bulgarian NLU | 73.19 | NLU Average Score (%) | 100 |
| EuroEval Latvian NLU - Latvian Twitter Sentiment | 56.55 | Sentiment classification Score (%) | 100 |
| EuroEval Polish NLU - KPWr NER | 78.64 | Named entity recognition Score (%) | 100 |
| EuroEval Slovene NLU - SSJ500K NER | 83.51 | Named entity recognition Score (%) | 100 |
| EuroEval Swedish Knowledge | 88.01 | Knowledge Average Score (%) | 100 |
| EuroEval Swedish NLU - SUC3 | 83.44 | Named entity recognition Score (%) | 100 |
| EuroEval Ukrainian NLU - NER UK | 83.05 | Named entity recognition Score (%) | 100 |
| EuroEval Portuguese NLU - ScaLA PT | 74.28 | Linguistic acceptability Score (%) | 99.8 |
| EuroEval German NLU - ScaLA DE | 72.72 | Linguistic acceptability Score (%) | 99.7 |
| EuroEval Czech NLU | 62.29 | NLU Average Score (%) | 99.6 |
| EuroEval Albanian NLU - MMS SQ | 30.83 | Sentiment classification Score (%) | 99.5 |
Interactive version: theaggregate.ai/model?slug=gpt-5-4-mini-2026-03-17 · How It Works · Data refreshed daily, snapshot 2026-10-09.