Kimi K2.6 is an open-source native multimodal agentic model built on a Mixture-of-Experts architecture with 1 trillion total parameters and 32 billion active parameters. It processes text, images, and video through the MoonViT 400M vision encoder, supporting both thinking and non-thinking modes. The model excels at long-horizon coding across Rust, Go, Python, frontend, and DevOps, with agentic orchestration capable of coordinating up to 300 sub-agents executing 4,000 steps. Its design emphasizes autonomous execution and self-correction for complex software engineering tasks.

Released under a Modified MIT License for commercial and non-commercial use, Kimi K2.6 demonstrates a breakthrough in sustained coding reliability across diverse languages and scenarios, with significantly improved instruction compliance and self-correction. The model targets developers and researchers requiring advanced multimodal agentic AI, positioning it as a forward-looking solution for complex software engineering and autonomous agent workflows.