Compare / GPT-5 vs GPT-5.6 Terra
GPT-5 vs GPT-5.6 Terra
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 | GPT-5.6 Terra |
|---|---|---|---|
| Arena (formerly LMArena)reported | OverallArena rating | 1,421 | 1,467 |
| Arena (formerly LMArena)reported | Business, management and financeArena rating | 1,413 | 1,466 |
| Arena (formerly LMArena)reported | Creative writingArena rating | 1,373 | 1,428 |
| Arena (formerly LMArena)reported | Expert promptsArena rating | 1,460 | 1,504 |
| Arena (formerly LMArena)reported | Instruction followingArena rating | 1,410 | 1,462 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,435 | 1,465 |
| Arena (formerly LMArena)reported | Writing, literature and languageArena rating | 1,397 | 1,445 |
| Artificial Analysisreported | Artificial Analysis Coding Indexindex score | 37.8 | 76.7 |
| Artificial Analysisreported | Artificial Analysis Intelligence Indexindex score | 23.0 | 42.1 |
| Artificial Analysisreported | GPQA Diamond% of questions | 85.4% | 92.5% |
| Artificial Analysisreported | Humanity's Last Exam% of questions | 28.5% | 42.9% |
| Artificial Analysisreported | IFBench% of instructions | 73.1% | 71.2% |
| Artificial Analysisreported | Long-context reasoning (AA-LCR)% of questions | 78.2% | 83.0% |
| Artificial Analysisreported | Terminal-Bench 2.1% of tasks | 35.2% | 88.0% |
| Artificial Analysisreported | Terminal-Bench Hard% of tasks | 37.9% | 62.9% |
| Artificial Analysisreported | Τ-bench banking% of tasks | 22.1% | 40.2% |
| Artificial Analysisreported | Τ²-bench telecom% of tasks | 86.5% | 86.3% |
| SimpleQA Verified (Epoch AI)reported | Correct answers% of questions | 50.1% | 43.2% |
| UGI Leaderboardreported | Requested-length error% off the requested word count | 58.0% | 22.0% |
| UGI Leaderboardreported | Style adherencescore from 0 to 1 | 0.39 | 0.39 |
| UGI Leaderboardreported | Writing scorescore out of 100 | 53.5 | 60.4 |
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.