Unverified90% confidenceFactTime unknown
Kimi K2.5本身的后训练算力仅占Composer 2.5总训练算力的7.5%
1
Sources
90%
Confidence
Medium-term (~90 days)
Relevance
5/31/2026
First Seen
Valid until: 8/29/2026
Sources
Cursor Composer 2.5深度解析：1/10成本比肩Claude Opus 4.7的编码能力
bilibili多模态实验室
Related Entities
Related Claims
UnverifiedKimi's own K2.6 version ranked only 13th on the same SWE-bench leaderboard, while Composer 2.5 trained on the open-source K2 ranked third.77% similarVerifiedKimi K2.7 Code的思维链token使用量相比K2.6减少约30%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% 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）65% 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%64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/12612API
curl https://kongchang.com/api/v1/knowledge/claims/12612MCP
get_claim(id=12612)