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

AI coding company Cursor launched a new model this week called Composer 2, which it promotes as offering “frontier-level cryptographic intelligence.”

However, an

As proof, Finn pointed to the code that seemed to identify Kimmy as a model.

“[A]At least rename the form ID,” they scoffed.

It was a surprising discovery, given that Cursor is a well-funded US startup that raised a $2.3 billion round last fall at a valuation of $29.3 billion and reportedly has annual revenues exceeding $2 billion. The company also didn’t mention anything about Moonshot AI or Kimi in its announcement.

However, Cursor’s vice president of developer education, Lee Robinson, was quick to admit, “Yes, Composer 2 started from an open source base!” But he said: “Only a quarter of the compute spent on the final model came from the base, and the rest from our training.” As a result, he said, Composer 2’s performance across various benchmarks is “quite different” from Kimi’s.

Robinson also insisted that Cursor’s use of Kimi was consistent with the terms of its license, a point that Kimi’s account on

“We are proud to see Kimi-k2.5 providing the foundation,” Kimi’s account said. “Seeing our model effectively integrated through continuous Cursor pre-training and high-compute RL training is the open model ecosystem we love supporting.”

So why don’t we acknowledge Kimmy up front? Aside from any potential embarrassment in not creating a model from scratch, building on the Chinese model may seem particularly risky right now, with the so-called AI “arms race” often framed as an existential battle between the US and China. (See, for example, the apparent panic that hit Silicon Valley after Chinese company DeepSeek launched a competing model early last year.)

“It was a mistake not to mention Kimi’s rule in our blog from the beginning. We will fix that in the next model,” admitted Aman Sanger, co-founder of Cursor.