GPT-4.1: benchmark results
OpenAI's GPT-4.1 model with improved coding, instruction following, and a 1M-token context. Provider: OpenAI. Released 2025-04-14. Access: API.
Aggregate Rating 1679 ± 3, rank #248 of 1656 rated models, from 2385 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AEGIS (Academic Image Forensics) - Forgery Scope Discrimination | 59.98 | Macro-F1 (%) over three classes, telling whether an image is | 100 |
| AGC-Bench - futuregen | 2.15 | Dataset z-score | 100 |
| ARFBench - Magnitude | 68.4 | Accuracy (%) on the 76 Magnitude questions (how far the anom | 100 |
| ArabicCulturalQA | 79.1 | MCQ Accuracy Avg (%) | 100 |
| ArgBench (Chain-of-Thought) | 57.7 | Macro average (0-1 F1-type scores, times 100) over the five | 100 |
| ArgBench (Chain-of-Thought) - Argument Mining | 54.2 | Macro F1 (0-1, times 100) averaged over the argument mining | 100 |
| ArgBench (Chain-of-Thought) - Argument Reasoning | 46.5 | Macro F1 (0-1, times 100) averaged over the argument reasoni | 100 |
| ArgBench (Chain-of-Thought) - Perspective Assessment | 60.3 | Macro F1 (0-1, times 100) averaged over the argument perspec | 100 |
| ArgBench (Few-Shot) | 63.3 | Macro average (0-1 F1-type scores, times 100) over the five | 100 |
| ArgBench (Few-Shot) - Argument Mining | 59.7 | Macro F1 (0-1, times 100) averaged over the argument mining | 100 |
| ArgBench (Few-Shot) - Argument Reasoning | 56 | Macro F1 (0-1, times 100) averaged over the argument reasoni | 100 |
| ArgBench (Few-Shot) - Perspective Assessment | 66.4 | Macro F1 (0-1, times 100) averaged over the argument perspec | 100 |
Interactive version: theaggregate.ai/model?slug=gpt-4-1 · How It Works · Data refreshed daily, snapshot 2026-10-20.