Kimi K2.7: benchmark results
Provider: Moonshot. Access: API.
Aggregate Rating 1706 ± 18, rank #187 of 1635 rated models, from 21 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| FirmCORe - Collaboration Type | 60.23 | Macro-F1 (%) | 100 |
| FirmCORe - Role Direction | 67.71 | Macro-F1 (%) | 100 |
| Sense Ruby Navigation - Baseline | 71.38 | Mean cited recall across repositories (%) | 75 |
| FirmCORe - Opportunity Detection | 52.87 | Macro-F1 (%) | 62.5 |
| Lean AI Formalization Leaderboard | 2 | Solved Problems | 48.6 |
| ASI-Bench (Claude Code) | 27.34 | Scientific Score (0-100) | 25 |
| ASI-Bench (Claude Code) - B1 Full Guidance | 44.43 | Scientific Score (0-100) | 25 |
| ASI-Bench (Claude Code) - B2 Method Hints | 23.84 | Scientific Score (0-100) | 25 |
| ASI-Bench (Claude Code) - B4 Goal + Distractors | 21.34 | Scientific Score (0-100) | 25 |
| ASI-Bench (Claude Code) - B3 Goal Only | 19.75 | Scientific Score (0-100) | 16.7 |
| DataBench | 12.7 | Score (%) | 7 |
| A2Z GameSpec-Bench - Adaptive Playtest Big | 26.2 | Adaptive Playtest Big (%; coding-agent game design fidelity; | 0 |
Interactive version: theaggregate.ai/model?slug=kimi-k2-7 · How It Works · Data refreshed daily, snapshot 2026-11-05.