Kimi K3 is Kimi's flagship multimodal model for long-horizon coding, knowledge work, and complex reasoning. It has 2.8 trillion parameters, native visual understanding, and a one-million-token context window. Built with architectural techniques including KDA hybrid linear attention and Attention Residuals, it can handle large codebases, long-running tasks, and coordinated tool use. It also supports image and video input, structured output, dynamic tool loading, and automatic context caching, making it well suited to complex software engineering, agent workflows, and demanding knowledge tasks.
Tabbit Browser has added Kimi K3 immediately. Pro members can select it from the model list and use it on both desktop and mobile.