Moonshot AI Releases Kimi, a Multimodal Coding Model
The new Mixture-of-Experts model from the Chinese AI company can generate code while also understanding visual inputs, a rare combination in open models.
Moonshot AI, the company behind the Kimi chatbot, has released a new open-weights model specialized for programming. The model, named Kimi-K2.7-Code, is now available for download on Hugging Face and brings a unique set of capabilities to the developer community.
Built on a Mixture-of-Experts (MoE) architecture, Kimi-K2.7-Code is designed for computational efficiency by only activating necessary parts of the network for any given task. While optimized for code generation and understanding, its most distinctive feature is multimodality. The model can process both text and visual inputs, a capability often reserved for larger, proprietary systems.
A Vision for Code Generation
The ability to see and interpret images sets Kimi-K2.7-Code apart from most other open coding models. This allows for powerful new workflows, such as generating front-end code from a UI mockup, creating data visualizations from a hand-drawn chart, or debugging visual artifacts in an application by providing a screenshot.
While Moonshot AI has not disclosed the model's specific parameter count or context length, its release marks a significant step in combining advanced architectures with practical, multimodal use cases for developers. The model is available under a custom "Kimi License Agreement," which users should review for specific usage terms.
Sources
More from Moonshot AI
All Moonshot AI releases →
Moonshot AI releases Kimi K3, a 2.8T-parameter MoE model
The open-weights multimodal model leans into coding and agentic tasks, extending Moonshot's Kimi line into a new scale bracket.
Kimi K2.6 tops closed models in coding test
Moonshot AI's open-weights mixture-of-experts model reportedly outperformed Claude, GPT-5.5, and Gemini on a programming challenge.
Moonshot AI Releases Kimi-K2.6 Multimodal Model
The Chinese AI lab has published weights for its new vision-language model, though a restrictive license limits its use to research applications.
More in Code
All Code →
MiniMax Debuts M3.1 Flash, a Multimodal MoE Model
The new release pairs a mixture-of-experts design with agentic, coding, and video-understanding abilities in a single open model.
Meta's Muse Spark Open Targets Agentic Coding
The new open-weight model pairs a million-token context window with Meta's bet on long-horizon coding agents.
Meta's Muse Spark Open Targets Agentic Coding
The open-weight model pairs a million-token context window with a focus on long-horizon code tasks.
0 comments
No comments yet. Be the first to weigh in.