Cursor has released Composer 2, its proprietary AI model for coding. The model performs significantly better than its predecessors on all benchmarks. A faster version is also available.
Composer 2 is available in the AI code editor. The model is the result of improved pretraining and reinforcement learning on long-running coding tasks. This enables Composer 2 to handle complex tasks that require hundreds of steps.
On three benchmarks, the model scores significantly better than its predecessors. On the proprietary CursorBench for measuring the performance of coding assistants, Composer 2 achieves a score of 61.3, compared to 44.2 for Composer 1.5 and 38.0 for Composer 1. Similar improvements are seen on Terminal-Bench 2.0 (a test for complex tasks within a CLI) and SWE-bench Multilingual (a test for solving software problems in multiple languages).
Frontier-level quality at a lower price
Composer 2 costs $0.50 per million input tokens and $2.50 per million output tokens. In addition to the standard version, there is a faster variant that is equivalent in intelligence but costs more: $1.50 per million input tokens and $7.50 per million output tokens. Cursor makes the fast variant the default option. According to the company, this price remains lower than that of other fast models on the market.
For users with an individual plan, Composer usage falls within a separate usage pool with a generous base rate. The model is now available in Cursor itself and in an early alpha version of a new interface.
Tip: Cursor bolsters AI coding platform with acquisition of Graphite