Cursor Integrates Kimi K3 Model for High-Performance Coding
Cursor has integrated the Kimi K3 model into its AI code editor, delivering near-frontier coding performance. The model is hosted on US-based inference platforms including Fireworks, Together, and Baseten, and supports zero data retention. This integration provides developers with access to a powerful 2.8-trillion-parameter open model optimized for long-horizon coding and complex reasoning tasks directly within their editor. By partnering with US-based inference providers, Cursor addresses enterprise privacy concerns through guaranteed zero data retention. Kimi K3 scores close to the frontier on CursorBench, Cursor's benchmark for evaluating multi-file coding-agent tasks. The model features a 1-million-token context window and native vision capabilities, enabling it to process massive codebases.
## BACKGROUND
Cursor is a popular AI-powered code editor designed to assist developers by automating coding tasks and answering codebase queries. Kimi K3 is a massive 2.8T-parameter open-weights model developed by Moonshot AI, utilizing advanced attention mechanisms for long-context reasoning. CursorBench is a benchmark created by Cursor to evaluate how well AI models perform on real-world, multi-file coding scenarios.