Cursor releases Composer2 technical report: RL environment fully simulates real user scenarios, base model score improves by 70%

CryptoWorld News reports that, according to 1M AI News monitoring, Cursor has released the Composer 2 technical report, revealing the complete training scheme for the first time. The base model Kimi K2.5 is built on MoE architecture, with a total of 1.04 trillion parameters and 32 billion activated parameters. The training is conducted in two phases: first, continued pretraining on code data to enhance encoding knowledge, then large-scale reinforcement learning to improve end-to-end coding capabilities. The RL environment fully simulates real Cursor usage scenarios, including file editing, terminal operations, code search, and tool calls, allowing the model to learn under conditions close to production environments. The report also details the construction of their self-developed benchmark CursorBench: tasks are collected from real coding sessions of engineering teams, rather than artificially created. The base Kimi K2.5 scores only 36.0 on this benchmark, but after two-phase training, Composer 2 reaches 61.3, a 70% improvement. Cursor states that its inference costs are significantly lower than cutting-edge models like GPT-5.4 and Claude Opus 4.6, achieving Pareto optimality between accuracy and cost.