GPT-4.1 Mini: benchmark results
OpenAI's mid-size GPT-4.1 tier with a 1M-token context, cheaper and lower-latency than GPT-4o at similar quality (April 2025). Provider: OpenAI. Released 2025-04-14. Access: API.
Aggregate Rating 1630 ± 3, rank #355 of 1656 rated models, from 2633 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AgentCollabBench - CLC | 2.6 | Cross-task Leakage Containment (%) | 100 |
| AgentCollabBench - CPR | 17.7 | Consensus Pollution Rate (%) | 100 |
| Appear2Meaning - Title | 54 | Title attribute accuracy (%, times 100; judged correct) on t | 100 |
| BlueBench - Chatbot Abilities | 97.55 | Score (%) | 100 |
| BlueBench - RAG General | 54.58 | Score (%) | 100 |
| CAME-Bench - Large | 21.3 | Answer-set macro F1 (%; 61 questions on two trajectories ave | 100 |
| ClosureBench (Chain-of-Thought) - L3 Composition | 74.9 | Accuracy (%; lenient scoring of one JSON answer per question | 100 |
| EuroEval Bosnian | 63.93 | Average Score (%) | 100 |
| EuroEval Bosnian NLU - MMS BS | 56.43 | Matthews correlation coefficient (%) | 100 |
| Galileo Agent - Telecom Accuracy | 64 | Accuracy (%) | 100 |
| HELM Robo-Reward-Bench - Berkeley Cable Routing | 0.65 | Absolute error | 100 |
| InteractWeb-Bench - Hallucination Rate | 23.5 | Hallucination rate (%): share of tasks where the agent build | 100 |
Interactive version: theaggregate.ai/model?slug=gpt-4-1-mini · How It Works · Data refreshed daily, snapshot 2026-10-20.