Compare / GPT-5.6 Terra vs GPT-5 Mini
GPT-5.6 Terra vs GPT-5 Mini
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-5.6 Terra | GPT-5 Mini |
|---|---|---|---|
| Arena (formerly LMArena)reported | OverallArena rating | 1,467 | 1,394 |
| Arena (formerly LMArena)reported | Business, management and financeArena rating | 1,466 | 1,381 |
| Arena (formerly LMArena)reported | Creative writingArena rating | 1,428 | 1,324 |
| Arena (formerly LMArena)reported | Expert promptsArena rating | 1,504 | 1,400 |
| Arena (formerly LMArena)reported | Instruction followingArena rating | 1,462 | 1,373 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,465 | 1,390 |
| Arena (formerly LMArena)reported | Writing, literature and languageArena rating | 1,445 | 1,352 |
| Artificial Analysisreported | Artificial Analysis Coding Indexindex score | 76.7 | 15.6 |
| Artificial Analysisreported | Artificial Analysis Intelligence Indexindex score | 42.1 | 20.6 |
| Artificial Analysisreported | GPQA Diamond% of questions | 92.5% | 82.8% |
| Artificial Analysisreported | Humanity's Last Exam% of questions | 42.9% | 21.5% |
| Artificial Analysisreported | IFBench% of instructions | 71.2% | 75.4% |
| Artificial Analysisreported | Long-context reasoning (AA-LCR)% of questions | 83.0% | 72.3% |
| Artificial Analysisreported | SciCode% of problems | 55.0% | 39.0% |
| Artificial Analysisreported | Terminal-Bench 2.1% of tasks | 88.0% | 3.7% |
| Artificial Analysisreported | Terminal-Bench 4.0% of tasks | 35.4% | 0.0% |
| Artificial Analysisreported | Terminal-Bench Hard% of tasks | 62.9% | 33.3% |
| Artificial Analysisreported | Τ-bench banking% of tasks | 40.2% | 15.5% |
| Artificial Analysisreported | Τ²-bench telecom% of tasks | 86.3% | 71.1% |
| SimpleQA Verified (Epoch AI)reported | Correct answers% of questions | 43.2% | 21.6% |
| Vending-Bench 2reported | Money after a yearUS dollars | $7,343.21 | $-31.18 |
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.