Cursor Admits: Kimi AI Powered Its New Coding Model

Cursor, an artificial intelligence coding firm, unveiled its latest model, Composer 2, this week, touting it as possessing “frontier-level coding intelligence.”

However, an individual on X, identified as Fynnsoon, alleged that Composer 2 was merely “just Kimi 2.5” supplemented with reinforcement learning. Kimi 2.5 is an open-source model recently introduced by Moonshot AI, a Chinese entity supported by Alibaba and HongShan (previously Sequoia China).

As substantiation for this claim, Fynn highlighted specific code snippets that appeared to identify Kimi as the underlying model.

“At least rename the model ID,” they remarked derisively.

This disclosure proved surprising, given Cursor’s status as a well-capitalized U.S. startup that secured a $2.3 billion funding round last autumn, achieving a $29.3 billion valuation. The company is also reportedly generating over $2 billion in annualized revenue. Furthermore, Cursor’s initial announcement made no mention of Moonshot AI or Kimi.

Nonetheless, Lee Robinson, Cursor’s vice president of developer education, promptly conceded, “Yep, Composer 2 started from an open-source base!” He further clarified, “Only about one-quarter of the computational resources invested in the final model originated from this base, with the remaining portion stemming from our proprietary training.” Consequently, he asserted that Composer 2’s performance across various benchmarks is “very different” from that of Kimi.

Robinson also maintained that Cursor’s utilization of Kimi adhered fully to the terms of its license. This assertion was corroborated by the official Kimi account on X in a subsequent post congratulating Cursor, which stated that Cursor employed Kimi “as part of an authorized commercial partnership” with Fireworks AI.

“We are proud to see Kimi-k2.5 provide the foundation,” the Kimi account declared. It added, “Seeing our model integrated effectively through Cursor’s continued pretraining & high-compute RL training is the open model ecosystem we love to support.”

The question then arises: why was Kimi not acknowledged initially? Beyond the potential for embarrassment stemming from not developing a model entirely from scratch, utilizing a Chinese-developed model could be perceived as particularly sensitive in the current climate. The ongoing AI “arms race” is frequently characterized as an existential confrontation between the United States and China, a sentiment exemplified by Silicon Valley’s apparent concern following the release of a competitive model by Chinese firm DeepSeek early last year.

Aman Sanger, co-founder of Cursor, conceded, “It was a miss to not mention the Kimi base in our blog from the start. We’ll fix that for the next model.”