Kimi K2.7 Code: benchmark results
Moonshot's open coding-focused Kimi K2.7 Code MoE model (1T total, 32B active). Provider: Moonshot. Released 2026-06-12. Access: Open.
Aggregate Rating 1772 ± 7, rank #89 of 1656 rated models, from 344 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| LLM Stats (MCP-Mark) | 81.1 | Score (%) | 100 |
| LiveClawBench | 76 | Average reward (0-100) over three runs on all 134 cases, Ope | 100 |
| LiveClawBench - Easy | 95.6 | Average reward (0-100) over three runs on the 69 Easy cases | 100 |
| LiveClawBench - Pass@3 | 70.9 | Pass@3 (x100): share of the 134 cases where at least one of | 100 |
| PACT | 0.94 | PACTScore (0-1) | 100 |
| PACT - Pushback Resistance | 0.99 | Score (0-1) | 100 |
| AA Omniscience - Software Engineering (SWE) - TypeScript | 66.67 | Accuracy (%) | 98.6 |
| AA Omniscience - Software Engineering (SWE) - PHP | 74 | Accuracy (%) | 98.5 |
| ReLE - Medicine and Mental Health | 90.8 | Accuracy (%) | 98.4 |
| AA Omniscience - Software Engineering (SWE) - C | 76 | Accuracy (%) | 98.3 |
| AA Omniscience - Software Engineering (SWE) - Kotlin | 54 | Accuracy (%) | 98.1 |
| AA Omniscience - Software Engineering (SWE) - Java | 43 | Accuracy (%) | 98 |
Interactive version: theaggregate.ai/model?slug=kimi-k2-7-code · How It Works · Data refreshed daily, snapshot 2026-10-20.