Cursor, the AI coding platform valued at $29.3 billion, has released Composer 2, a new in-house coding model that dramatically improves on its predecessor while slashing costs by up to 86%.

Built as a fine-tuned variant of China's open-source Kimi K2.5 model, Composer 2 is available exclusively within Cursor's coding environment at $0.50/$2.50 per million input/output tokens—down from Composer 1.5's $3.50/$17.50 pricing. A faster variant, Composer 2 Fast, costs $1.50/$7.50 and is now the default experience.

The model shows substantial benchmark improvements, scoring 61.3 on CursorBench and 73.7 on SWE-bench Multilingual, compared to Composer 1.5's 44.2 and 65.9 respectively. On Terminal-Bench 2.0, Composer 2 scores 61.7—beating Claude Opus 4.6's 58.0 but trailing GPT-5.4's leading 75.1.

Cursor emphasizes that Composer 2 is optimized for "long-horizon agentic coding"—complex workflows requiring hundreds of actions across multiple files, terminal commands, and debugging cycles. The model features a 200,000-token context window and tight integration with Cursor's tool stack.

However, the release comes as Cursor faces growing competitive pressure from first-party coding products like Claude Code and OpenAI's Codex. Social media chatter suggests some power users are migrating to these direct alternatives, questioning whether Cursor's platform adds sufficient value between developers and model makers.