BREAKING
Moonshot Open-Sources Kimi-K2.7-Code
Coding scores jump over K2.6
Kimi Code Bench v2
62
Program Bench
53.6
K2.6 Code Bench
50.9
0
%
fewer reasoning tokens
0
%
Kimi Code Bench v2 gain
0
%
MLS Bench Lite gain
Open and cheap vs closed models
K2.7-Code
Open
●
Modified MIT license
●
Output $4.00 per 1M tokens
●
256K context, Native INT4
Closed leaders
Benchmark
●
GPT-5.5 scores 69.0
●
Claude Opus 4.8 scores 67.4
●
K2.7-Code scores 62.0
Available now, with caveats
1
On Hugging Face
↓
2
Kimi API and Code CLI
↓
3
In-house benchmarks only
↓
4
Third-party check needed
Coding focus drives K2 series forward
AI NEWS BLITZ
Moonshot AI has released its coding-focused Kimi-K2.7-Code as open source.