“An analysis found that Cursor's Composer 2.5 coding model performs at a comparable level to Anthropic's Opus 4.7 and OpenAI's GPT-5.5.”