| OpenAI o4-mini | OpenAI o3 | |
|---|---|---|
| arenaElo | 1362 | 1418 |
| humanEval | 88.6 | 92.4 |
| sweBench | 60 | 69.1 |
| mmlu | 85.4 | 90.2 |
| gpqa | 74.2 | 83.3 |
| aime | 92 | 91.6 |
| Price in/out ($/M) | $1.1 / $4.4 | $10 / $40 |
o4-mini lands within 10 GPQA points of o3 at ~10% of the price. Unless you need the absolute top of the reasoning curve, o4-mini is the smarter default for reasoning workloads at scale.