Up to four models
GPT-5 mini vs GPT-4.1 mini
On the index (GPT-4 = 100), GPT-5 mini scores 140 and GPT-4.1 mini 120. Side by side on the same third-party tests, with price and context. The link in your address bar shares this exact comparison.
GPT-5 mini
140124–156
WeightsClosed
ReleasedAug 2025
Price in / out$0.25 / $2.0
Context400K
Core tests taken5 of 9
GPT-4.1 mini
120104–136
WeightsClosed
ReleasedApr 2025
Price in / out$0.40 / $1.6
Context1.05M
Core tests taken3 of 9
GPT-5 mini leads GPT-4.1 mini by 20 points, ahead on 3 of 3 shared core tests, at a similar price.
Test by testthird-party results
GPT-5 miniGPT-4.1 mini
Every result as a table
| Test | GPT-5 mini | GPT-4.1 mini |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 75.0% | 65.8% |
| GPQA Diamond (Vals AI) | 80.3% | 67.9% |
| MMLU Pro | 82.2% | 77.2% |
| MMMU Pro | 78.9% | 70.5% |
| Fiction.LiveBench | 69.4% | 44.4% |
| SimpleQA Verified | 21.6% | 12.7% |
| Maths | ||
| FrontierMath (tiers 1-3) | 46.7% | 6.7% |
| AIME-style maths (OTIS mock) | 86.7% | 44.7% |
| MATH (level 5) | 97.9% | 87.3% |
| MATH 500 | 94.8% | 88.0% |
| MGSM | 92.6% | 87.8% |
| Chess puzzles | 24.0% | 1.0% |
| FrontierMath (2025 set) | 27.2% | 4.5% |
| Coding | ||
| LiveCodeBench | 86.6% | 58.2% |
| WeirdML | 52.7% | 37.6% |
| Agents | ||
| DTBench | 80.5% | 68.8% |
| LMCA | 34.2% | 21.1% |
| Mystery game puzzles | 10.0% | 7.0% |
| Professional work | ||
| CorpFin v2 | 60.2% | 57.9% |
| LegalBench | 81.8% | 78.0% |
| MortgageTax | 66.9% | 65.5% |
| TaxEval v2 | 75.2% | 71.9% |
| Novel problems | ||
| ARC-AGI-2 | 4.4% | 0.0% |
| ARC-AGI-1 | 54.3% | 3.5% |
Tests at least two of these models have taken. Bold rows are the core tests.