Cursor Admits: New AI Coding Model Built on Chinese Open Source

Cursor, the AI coding tool with $2 billion in annualized revenue and a nearly $30 billion valuation, has admitted its newest model Composer 2 is built on top of Kimi k2.5, an open source model from Chinese company Moonshot AI.

When Cursor launched Composer 2 this week, marketing it as offering "frontier-level coding intelligence," it was quickly exposed by an X user who pointed to code identifying Kimi as the underlying model. "At least rename the model ID," the user scoffed.

Cursor's VP of developer education Lee Robinson quickly confirmed the revelation, but stressed that significant training compute had been stacked on top of the base. "Only about a quarter of the compute spent on the final model came from the base," he explained, adding that Composer 2's benchmark performance is therefore "very different" from Kimi's.

Moonshot AI confirmed Cursor used Kimi "as part of an authorized commercial partnership" via Fireworks AI, welcoming the news and saying they were "proud to see Kimi-k2.5 provide the foundation."

The story raises interesting questions about what actually happens behind the scenes in the AI industry. Cursor is a heavily funded US startup, yet chose to build its newest model on a Chinese open source base without mentioning this in its launch announcement. That's legal, but the lack of transparency leaves a sour taste for users paying premium prices for what they believed was a unique Cursor product.

For technology leaders evaluating AI coding tools, the lesson is clear: behind the polish of AI products often lie complex supply chains and licensing arrangements. Using open source as a foundation is not unethical, but it should be communicated clearly to customers.