Compare / Grok 4.3 vs GLM 4.6
Grok 4.3 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 | Grok 4.3 | GLM 4.6 |
|---|---|---|---|
| Arena (formerly LMArena)reported | OverallArena rating | 1,428 | 1,447 |
| Arena (formerly LMArena)reported | Business, management and financeArena rating | 1,440 | 1,412 |
| Arena (formerly LMArena)reported | Creative writingArena rating | 1,423 | 1,400 |
| Arena (formerly LMArena)reported | Expert promptsArena rating | 1,452 | 1,439 |
| Arena (formerly LMArena)reported | Instruction followingArena rating | 1,415 | 1,412 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,442 | 1,424 |
| Arena (formerly LMArena)reported | Writing, literature and languageArena rating | 1,424 | 1,405 |
| Artificial Analysisreported | Artificial Analysis Coding Indexindex score | 42.2 | 45.8 |
| Artificial Analysisreported | Artificial Analysis Intelligence Indexindex score | 24.9 | 18.5 |
| Artificial Analysisreported | GPQA Diamond% of questions | 90.1% | 78.0% |
| Artificial Analysisreported | Humanity's Last Exam% of questions | 37.2% | 14.5% |
| Artificial Analysisreported | IFBench% of instructions | 83.3% | 43.4% |
| Artificial Analysisreported | Long-context reasoning (AA-LCR)% of questions | 75.0% | 54.0% |
| Artificial Analysisreported | Terminal-Bench 2.1% of tasks | 39.7% | 49.4% |
| Artificial Analysisreported | Terminal-Bench Hard% of tasks | 37.9% | 28.8% |
| Artificial Analysisreported | Τ-bench banking% of tasks | 12.4% | 13.4% |
| Artificial Analysisreported | Τ²-bench telecom% of tasks | 97.7% | 76.9% |
| UGI Leaderboardreported | Requested-length error% off the requested word count | 21.0% | 13.0% |
| UGI Leaderboardreported | Style adherencescore from 0 to 1 | 0.34 | 0.38 |
| UGI Leaderboardreported | Writing scorescore out of 100 | 57.7 | 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.