Kimi.ai releases Kimi-K2.7-Code as an open coding model
Kimi.ai says Kimi-K2.7-Code beats K2.6 on coding benchmarks and is meant for Kimi Code and the Kimi API, while a new beta program will give applicants early access to upcoming models and features.
By Ryan Merket · Published
Why it matters
Open coding models are becoming distribution plays: Kimi.ai is using weights, API access, and a beta program to pull developers into its own coding workflow before the next model cycle.
Kimi.ai (@Kimi_Moonshot) released Kimi-K2.7-Code on Friday, describing it in a four-post thread on X as its latest coding model and saying the model is open-sourced.
The concrete release artifact is a Hugging Face repository for Kimi-K2.7-Code, which Kimi.ai linked as the location for the model's weights and code. The company is positioning the model specifically for coding and agent work, not as a general-purpose replacement for K2.6.
Kimi.ai claims Kimi-K2.7-Code improves over K2.6 by 21.8% on Kimi Code Bench v2, 11.0% on Program Bench, and 31.5% on MLS Bench Lite. It also claims 30% lower reasoning overhead.
Kimi.ai also said it is launching the Kimi Code Beta Program, inviting users to apply to try upcoming models and features before public release.