Improved reasoning over challenging coding tasks by scaling RL over 20x.

Cursor

Composer 1.5 is an upgraded agentic coding model that uses 20x more reinforcement learning than its predecessor. It balances speed and intelligence by adapting its thinking time based on problem difficulty, generates reasoning tokens for complex tasks, and includes self-summarization capabilities to handle longer contexts. The model shows continued performance improvements on real-world coding benchmarks, especially on challenging problems.

Introducing Composer 1.5