GPT-4.1 vs GPT-5 Nano
Lado a lado. Cada métrica. Cada benchmark.
| Tipo | GPT-4.1 | GPT-5 Nano |
|---|---|---|
| Provider |  OpenAI | OpenAI |
| pontuação média | 43.3 | 45.3 |
| Preço de entrada | $2.00 | $0.05 |
| Preço de saída | $8.00 | $0.40 |
| Janela de contexto | 1.0M tokens (~524 books) | 400K tokens (~200 books) |
| Lançado em | 2025-04-14 | 2025-08-07 |
| Código aberto | Proprietary | Proprietary |
Pontuações de benchmark
15 benchmarks · GPT-4.1: 5, GPT-5 Nano: 10
| Benchmark | Categoria | GPT-4.1 | GPT-5 Nano |
|---|---|---|---|
| ARC-AGI | reasoning | 5.5 | 20.7 |
| ARC-AGI-2 | reasoning | 0.4 | 2.6 |
| Fiction.LiveBench | knowledge | 63.9 | 44.4 |
| FrontierMath-2025-02-28-Private | math | 5.5 | 8.3 |
| FrontierMath-Tier-4-2025-07-01-Private | math | 0.1 | 2.1 |
| GPQA diamond | knowledge | 55.9 | 59.3 |
| HELM — GPQA | knowledge | 65.9 | 67.9 |
| HELM — IFEval | language | 83.8 | 93.2 |
| HELM — MMLU-Pro | knowledge | 81.1 | 77.8 |
| HELM — Omni-MATH | math | 47.1 | 54.7 |
| HELM — WildBench | reasoning | 85.4 | 80.6 |
| MATH level 5 | math | 83.0 | 95.2 |
| OTIS Mock AIME 2024-2025 | math | 38.3 | 81.1 |
| SWE-Bench Verified (Bash Only) | coding | 39.6 | 34.8 |
| WeirdML | coding | 39.0 | 38.1 |