OpenAI
GPT-4.1 mini benchmark results
11 results on 8 benchmarks from 5 sources.
Also listed as: gpt-4.1-mini-2025-04-14 · gpt-4.1-mini-20250414 · GPT 4.1 mini · GPT 4.1 mini (2025-04-14) · GPT-4.1-mini-2025-04-14 (FC) · GPT-4.1-mini-2025-04-14 (Prompt) · GPT-4.1-mini
All benchmark results
| Benchmarks | Score | Rank | Setting | Source type | Date |
|---|---|---|---|---|---|
| LMArena Text | 1383 | 185 | Overall, style control on; 38709 votes; 95% CI 1378–1387“gpt-4.1-mini-2025-04-14” | Official leaderboard | Oct 8, 2026 |
| LMArena Vision | 1201 | 84 | Overall, style control on; 40223 votes; 95% CI 1193–1208“gpt-4.1-mini-2025-04-14” | Official leaderboard | Oct 2, 2026 |
| GPQA Diamond | 65.8% | 173 | Epoch AI run, accuracy | Independent run | Apr 14, 2025 |
| SWE-bench Verified | 23.9% | 96 | agent: mini-SWE-agent“gpt-4.1-mini-20250414” | Official leaderboard | Jul 20, 2025 |
| AIME (OTIS Mock AIME 2024–2025) | 44.7% | 177 | Epoch AI run, OTIS Mock AIME 2024–2025 | Independent run | Apr 14, 2025 |
| MATH Level 5 | 87.3% | 26 | Epoch AI run, accuracy | Independent run | Apr 14, 2025 |
| τ²-bench | 48.7% | 25 | airline domain, pass^1; submitted by Sierra; v0.1.3“GPT-4.1-mini” | Official leaderboard | Jun 9, 2025 |
| τ²-bench | 61.4% | 26 | retail domain, pass^1; submitted by Sierra; v0.1.3“GPT-4.1-mini” | Official leaderboard | Jun 9, 2025 |
| τ²-bench | 48.9% | 20 | telecom domain, pass^1; submitted by Sierra; v0.1.3“GPT-4.1-mini” | Official leaderboard | Jun 9, 2025 |
| BFCL v4 (function calling) | 50.5% | 27 | Overall accuracy; native function calling (FC)“GPT-4.1-mini-2025-04-14 (FC)” | Official leaderboard | Oct 17, 2026 |
| BFCL v4 (function calling) | 29.7% | 67 | Overall accuracy; prompt-based“GPT-4.1-mini-2025-04-14 (Prompt)” | Official leaderboard | Oct 17, 2026 |
FAQ
Where do these GPT-4.1 mini scores come from?
From Epoch AI Benchmarking Hub, Berkeley Function Calling Leaderboard (BFCL), LMArena leaderboard dataset, SWE-bench leaderboards (swebench.com), and τ²-bench leaderboard submissions. Each row links to the original leaderboard, dataset or report, and vendor-reported numbers are labelled.