Cursor has unveiled the Composer 2 technical report, revealing a 70% improvement in performance on its proprietary benchmark, CursorBench. The model, Kimi K2.5, utilizes a MoE architecture with 1.04 trillion parameters and 32 billion activated parameters. Training involves two stages: pre-training on code data and reinforcement learning in simulated real-world environments. Composer 2 scored 61.3 on CursorBench, up from Kimi K2.5's initial 36.0. Cursor highlights its model's cost-effectiveness compared to leading APIs like GPT-5.4 and Claude Opus 4.6.
Cursor's Composer 2 Achieves 70% Improvement on Custom Benchmark
Disclaimer: The content provided on Phemex News is for informational purposes only. We do not guarantee the quality, accuracy, or completeness of the information sourced from third-party articles. The content on this page does not constitute financial or investment advice. We strongly encourage you to conduct you own research and consult with a qualified financial advisor before making any investment decisions.
