Cursor admits its new coding model was built on top of Moonshot AI’s Kimi

This week, AI coding firm Cursor introduced its latest model, Composer 2, boasting what it described as "frontier-level coding intelligence." However, the announcement was quickly met with skepticism when a user known as Fynn on X alleged that Composer 2 is essentially an enhanced version of Kimi 2.5, an open-source model developed by Moonshot AI, a Chinese company supported by Alibaba and HongShan (formerly Sequoia China). Fynn cited code that appeared to confirm Kimi's influence, even criticizing Cursor for not updating the model's ID. This revelation raised eyebrows given that Cursor is a prominent U.S. startup, having secured $2.3 billion in funding last fall and boasting a valuation of $29.3 billion, alongside reported annual revenues exceeding $2 billion. In response to the controversy, Cursor's vice president of developer education, Lee Robinson, admitted, "Yes, Composer 2 started from an open-source base!" However, he emphasized that only about 25% of the computational resources used in the final model came from Kimi, asserting that the performance benchmarks for Composer 2 differ significantly from those of Kimi. Robinson also maintained that Cursor's utilization of Kimi adhered to licensing agreements. This sentiment was echoed by the Kimi account on X, which congratulated Cursor and clarified that their collaboration was part of an authorized commercial partnership with Fireworks AI. "We take pride in seeing Kimi-k2.5 serve as a foundational element," the account stated, highlighting the effective integration of their model through Cursor's rigorous training processes. The question remains: why didn't Cursor acknowledge Kimi in its initial announcement? Building on a Chinese model could be seen as contentious, particularly in the current climate of an AI "arms race" between the U.S. and China. Cursor co-founder Aman Sanger recognized the oversight, stating, "It was a miss to not mention the Kimi base in our blog from the start. We'll address that in our future communications."