Expired50% confidenceFactExact time
Cursor used RLHF (Reinforcement Learning from Human Feedback) training on the open-source Kimi K2 model using data from real user programming interactions to produce Composer 2.5.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
Valid until: 9/30/2026(expired)
Sources
Cursor Composer 2.5: The Secret Behind an Open-Source Model's Reinforcement Training to a Top-3 Coding Benchmark Ranking
bilibili鲲鹏Talk5/20/2026
Related Claims
UnverifiedCursor uses implicit feedback from users in the IDE—such as accepting or rejecting code suggestions and modifying AI-generated code—as RLHF training signals to perform reinforcement training on Kimi K2.78% similarVerifiedThe Cursor team officially released its self-developed programming model Composer 2.5, trained on the Kimi K2.5 open-source checkpoint.73% similarUnverifiedCursor's Composer 2 is trained based on Kimi 2.5, which is a Mixture-of-Experts model with 1 trillion parameters and 30 billion active parameters.72% similarUnverifiedCursor performed reinforcement learning fine-tuning on top of Kimi KR, specializing the model through training on large amounts of high-quality programming task data.70% similarUnverifiedCursor partnered with Fireworks to build the distributed reinforcement learning infrastructure for Composer 2.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/43941API
curl https://kongchang.com/api/v1/knowledge/claims/43941MCP
get_claim(id=43941)