Unverified50% confidenceFactExact time
K2.7 Code目前尚未公布SWE-Bench成绩，所有性能数据均为月之暗面自报，此前K2.6曾出现自报成绩与第三方复现相差15-20个百分点的情况
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
Valid until: 9/30/2026
Sources
Kimi K2.7 Code深度解析：万亿参数编程模型的性价比之战
bilibiliAI睡前简报6/13/2026
Related Claims
UnverifiedK2.7 Code在MLS Bench Lite基准测试上相比K2.6提升31.5%（从26.7提升至35.1），是所有测试中提升幅度最大的70% similarPartially Verified与K2.6相比，K2.7 Code在Kimi Code Bench VR基准测试上提升21.8%（从50.9提升至62.0）70% similarUnverifiedKimi K2.7 Code在MCP Mark Verified基准测试上得分81.1，相比K2.6的72.8提升11.4%69% similarVerified在Terminal Bench 2.0测试中，K2.6打出66.7分，比GPT-5.4和Claude Opus 4.6都略高68% similarUnverifiedKimi K2.6在SWE-Bench Pro测试中获得58.6%的分数，位居开源模型第一66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/42607API
curl https://kongchang.com/api/v1/knowledge/claims/42607MCP
get_claim(id=42607)