Compare / Kimi K2.5 vs GPT-5.6 Luna
Kimi K2.5 vs GPT-5.6 Luna
20 results from 5 sources that measured both models. Each row is in the source's own units; there is no overall winner.
| Benchmark | Metric | Kimi K2.5 | GPT-5.6 Luna |
|---|---|---|---|
| Arena (formerly LMArena)reported | OverallArena rating | 1,436 | 1,462 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,445 | 1,461 |
| Arena (formerly LMArena)reported | Business, management and financeArena rating | 1,441 | 1,463 |
| Arena (formerly LMArena)reported | Creative writingArena rating | 1,423 | 1,412 |
| Arena (formerly LMArena)reported | Expert promptsArena rating | 1,483 | 1,491 |
| Arena (formerly LMArena)reported | Instruction followingArena rating | 1,439 | 1,446 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,450 | 1,454 |
| Arena (formerly LMArena)reported | Writing, literature and languageArena rating | 1,428 | 1,435 |
| Artificial Analysisreported | Artificial Analysis Coding Indexindex score | 46.8 | 71.4 |
| Artificial Analysisreported | Artificial Analysis Intelligence Indexindex score | 23.5 | 37.3 |
| Artificial Analysisreported | GPQA Diamond% of questions | 87.9% | 91.1% |
| Artificial Analysisreported | Humanity's Last Exam% of questions | 30.7% | 39.5% |
| Artificial Analysisreported | Long-context reasoning (AA-LCR)% of questions | 78.0% | 83.7% |
| Artificial Analysisreported | Terminal-Bench 2.1% of tasks | 45.7% | 80.9% |
| Artificial Analysisreported | Τ-bench banking% of tasks | 14.2% | 31.1% |
| SimpleQA Verified (Epoch AI)reported | Correct answers% of questions | 34.3% | 41.0% |
| UGI Leaderboardreported | Requested-length error% off the requested word count | 10.0% | 15.0% |
| UGI Leaderboardreported | Style adherencescore from 0 to 1 | 0.34 | 0.38 |
| UGI Leaderboardreported | Writing scorescore out of 100 | 62.4 | 53.7 |
| Vending-Bench 2reported | Money after a yearUS dollars | $1,198.46 | $4,094.71 |
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.