Compare / GPT-5 Mini vs Grok 4.3
GPT-5 Mini vs Grok 4.3
21 results from 4 sources that measured both models. Each row is in the source's own units; there is no overall winner.
| Benchmark | Metric | GPT-5 Mini | Grok 4.3 |
|---|---|---|---|
| Arena (formerly LMArena)reported | OverallArena rating | 1,394 | 1,428 |
| Arena (formerly LMArena)reported | Business, management and financeArena rating | 1,381 | 1,440 |
| Arena (formerly LMArena)reported | Creative writingArena rating | 1,324 | 1,423 |
| Arena (formerly LMArena)reported | Expert promptsArena rating | 1,400 | 1,452 |
| Arena (formerly LMArena)reported | Instruction followingArena rating | 1,373 | 1,415 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,390 | 1,442 |
| Arena (formerly LMArena)reported | Writing, literature and languageArena rating | 1,352 | 1,424 |
| Artificial Analysisreported | Artificial Analysis Coding Indexindex score | 15.6 | 42.2 |
| Artificial Analysisreported | Artificial Analysis Intelligence Indexindex score | 20.6 | 24.9 |
| Artificial Analysisreported | GPQA Diamond% of questions | 82.8% | 90.1% |
| Artificial Analysisreported | Humanity's Last Exam% of questions | 21.5% | 37.2% |
| Artificial Analysisreported | IFBench% of instructions | 75.4% | 83.3% |
| Artificial Analysisreported | Long-context reasoning (AA-LCR)% of questions | 72.3% | 75.0% |
| Artificial Analysisreported | SciCode% of problems | 39.0% | 48.3% |
| Artificial Analysisreported | Terminal-Bench 2.1% of tasks | 3.7% | 39.7% |
| Artificial Analysisreported | Terminal-Bench 4.0% of tasks | 0.0% | 0.0% |
| Artificial Analysisreported | Terminal-Bench Hard% of tasks | 33.3% | 37.9% |
| Artificial Analysisreported | Τ-bench banking% of tasks | 15.5% | 12.4% |
| Artificial Analysisreported | Τ²-bench telecom% of tasks | 71.1% | 97.7% |
| SimpleQA Verified (Epoch AI)reported | Correct answers% of questions | 21.6% | 33.2% |
| Vending-Bench 2reported | Money after a yearUS dollars | $-31.18 | $35.26 |
Bold green marks the better value on that metric. Where a source reports ranges that overlap, the difference may not be meaningful; see the benchmark page for ranges.