Cursor, the AI coding startup valued at $29.3 billion, launched a new model this week called Composer 2, promoting it as “frontier-level coding intelligence.” It did not mention that the model was built on top of Kimi 2.5, an open-source model from Moonshot AI, a Chinese company backed by Alibaba and HongShan.
Not long ago, an X user going by the name Fynn made a bold claim that Composer 2 was really just “Kimi 2.5” with some extra reinforcement learning thrown in. For context, Kimi 2.5 is an open-source model that was recently launched by Moonshot AI, a Chinese company that has the backing of Alibaba and HongShan (which used to be known as Sequoia China).
To back up their assertion, Fynn pointed out some code that appeared to label Kimi as the model in question.
“[A]t least rename the model ID,” they quipped.
The discovery was embarrassing for Cursor, which was founded in 2022 by four MIT graduates and has become one of the fastest-growing startups in the developer tools space. Its AI-native code editor, built as a fork of Visual Studio Code, reportedly surpassed $2 billion in annualised revenue earlier this year after a $2.3 billion raise last fall.
Cursor VP Lee Robinson acknowledged the base model, saying only about a quarter of Composer 2’s compute came from Kimi, with the rest from Cursor’s own reinforcement learning training. He said the final model’s benchmark performance is “very different” from Kimi’s. Moonshot AI confirmed the use was part of an authorized commercial partnership through Fireworks AI and publicly congratulated Cursor.
Co-founder Aman Sanger conceded the lack of disclosure was “a miss” and said they would credit the base model in future launches.
“We are proud to see Kimi-k2.5 provide the foundation,” the Kimi account said. “Seeing our model integrated effectively through Cursor’s continued pretraining & high-compute RL training is the open model ecosystem we love to support.”
The episode highlights two growing tensions in AI: the blurring line between building a model and fine-tuning someone else’s, and the quiet dependence of well-funded US AI companies on Chinese open-source infrastructure, a dependency that is widespread but rarely acknowledged publicly.
