Compare / GPT-5.6 Terra vs GLM 4.6
GPT-5.6 Terra vs GLM 4.6
20 results from 3 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 | GLM 4.6 |
|---|---|---|---|
| Arena (formerly LMArena)reported | OverallArena rating | 1,467 | 1,447 |
| Arena (formerly LMArena)reported | Business, management and financeArena rating | 1,466 | 1,412 |
| Arena (formerly LMArena)reported | Creative writingArena rating | 1,428 | 1,400 |
| Arena (formerly LMArena)reported | Expert promptsArena rating | 1,504 | 1,439 |
| Arena (formerly LMArena)reported | Instruction followingArena rating | 1,462 | 1,412 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,465 | 1,424 |
| Arena (formerly LMArena)reported | Writing, literature and languageArena rating | 1,445 | 1,405 |
| Artificial Analysisreported | Artificial Analysis Coding Indexindex score | 76.7 | 45.8 |
| Artificial Analysisreported | Artificial Analysis Intelligence Indexindex score | 42.1 | 18.5 |
| Artificial Analysisreported | GPQA Diamond% of questions | 92.5% | 78.0% |
| Artificial Analysisreported | Humanity's Last Exam% of questions | 42.9% | 14.5% |
| Artificial Analysisreported | IFBench% of instructions | 71.2% | 43.4% |
| Artificial Analysisreported | Long-context reasoning (AA-LCR)% of questions | 83.0% | 54.0% |
| Artificial Analysisreported | Terminal-Bench 2.1% of tasks | 88.0% | 49.4% |
| Artificial Analysisreported | Terminal-Bench Hard% of tasks | 62.9% | 28.8% |
| Artificial Analysisreported | Τ-bench banking% of tasks | 40.2% | 13.4% |
| Artificial Analysisreported | Τ²-bench telecom% of tasks | 86.3% | 76.9% |
| UGI Leaderboardreported | Requested-length error% off the requested word count | 22.0% | 13.0% |
| UGI Leaderboardreported | Style adherencescore from 0 to 1 | 0.39 | 0.38 |
| UGI Leaderboardreported | Writing scorescore out of 100 | 60.4 | 53.3 |
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.