BREAKING
Kimi-K2.7-Code hits No.3 open model
0
open-model rank
0
overall rank
0T
total params
0B
activated/token
0K
context length
Benchmark gains over K2.6
Kimi Code v2 K2.650.9
Kimi Code v2 K2.762
MCP Mark K2.672.8
MCP Mark K2.781.1
Strengths vs open questions
Strengths
~30% fewer thinking tokens
Stable on long coding loops
Cost-advantaged open weights
Caution
Benchmarks lean internal
Regressions on KernelBench
Awaiting verification
Open weights challenge closed leaders
AI NEWS BLITZ
Moonshot AI's new open-weight coding model just broke into the upper ranks of a major frontend leaderboard.