GPT-5.4 Nano (High): benchmark results
GPT-5.4 Nano evaluated at the high reasoning-effort setting. Provider: OpenAI. Released 2026-03-17. Access: API.
Aggregate Rating 1667 ± 18, rank #467 of 2168 rated models, from 90 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Prim - Discovery | 41.76 | Discovery score (%): how well the model states the problem's | 81.8 |
| InfiniteBM Heads-Up No-Limit Hold'em | 1282.53 | Arena Elo (elo) | 72.7 |
| InfiniteBM Liar's Dice | 1304.64 | Arena Elo (elo) | 72.7 |
| Chatbot Arena (Text - Math) | 1425 | Arena Score | 70.6 |
| ALE-Bench | 1004.52 | Performance (Self-Refine x1) (imported) | 70 |
| Chatbot Arena (Text - Mathematical) | 1432 | Arena Score | 69.3 |
| FrontierMath - Tiers 1-3 | 25.86 | Accuracy (%, 290 problems) | 67.7 |
| FWBench | 2.64 | Decision score S (lower is better; 0 is the hindsight-optima | 66.7 |
| ReLE - Reasoning and Mathematics | 77.2 | Accuracy (%) | 66.7 |
| Chatbot Arena (Text - Coding) | 1460 | Arena Score | 66.3 |
| Chatbot Arena (Text - Software & IT Services) | 1450 | Arena Score | 66.3 |
| ReLE - Agents and Tool Use | 63.1 | Accuracy (%) | 66.2 |
Interactive version: theaggregate.ai/model?slug=gpt-5-4-nano-high · How It Works · Data refreshed daily, snapshot 2026-10-20.