Compare / Claude Opus 4.7 vs Kimi K3
Claude Opus 4.7 vs Kimi K3
20 results from 7 sources that measured both models. Each row is in the source's own units; there is no overall winner.
| Benchmark | Metric | Claude Opus 4.7 | Kimi K3 |
|---|---|---|---|
| APEX-Agentsreported | Tasks passed% of tasks | 49.2% | 50.6% |
| Arena (formerly LMArena)reported | OverallArena rating | 1,477 | 1,472 |
| Arena (formerly LMArena)reported | Business, management and financeArena rating | 1,508 | 1,489 |
| Arena (formerly LMArena)reported | Creative writingArena rating | 1,488 | 1,458 |
| Arena (formerly LMArena)reported | Expert promptsArena rating | 1,535 | 1,533 |
| Arena (formerly LMArena)reported | Instruction followingArena rating | 1,503 | 1,486 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,502 | 1,488 |
| Arena (formerly LMArena)reported | Writing, literature and languageArena rating | 1,496 | 1,475 |
| Artificial Analysisreported | Artificial Analysis Coding Indexindex score | 73.6 | 76.2 |
| Artificial Analysisreported | Artificial Analysis Intelligence Indexindex score | 40.7 | 43.6 |
| Artificial Analysisreported | GPQA Diamond% of questions | 91.4% | 93.5% |
| Artificial Analysisreported | Humanity's Last Exam% of questions | 42.3% | 46.9% |
| Artificial Analysisreported | Long-context reasoning (AA-LCR)% of questions | 78.7% | 88.7% |
| Artificial Analysisreported | Terminal-Bench 2.1% of tasks | 83.1% | 85.0% |
| Artificial Analysisreported | Τ-bench banking% of tasks | 34.6% | 46.0% |
| GDP.pdfreported | Professional document tasks% of rubric | 21.0% | 19.0% |
| SimpleQA Verified (Epoch AI)reported | Correct answers% of questions | 51.7% | 50.6% |
| tau2-benchreported | Consistency (pass^4)% of tasks | 24.7% | 17.5% |
| tau2-benchreported | Task success (pass^1)% of tasks | 40.2% | 37.1% |
| Vending-Bench 2reported | Money after a yearUS dollars | $10,936.76 | $5,165.03 |
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.