Cursor, a San Francisco-based AI startup, has launched Composer 2, its latest in-house coding model. The new model offers improved benchmarks, lower pricing, and faster performance compared to its predecessor, Composer 1.5.
Composer 2 is designed specifically for Cursor’s agentic AI coding environment, which allows users to complete complex coding tasks with long-horizon coding capabilities. The model has been trained on a wide range of tasks, including file edits, terminal operations, and web access.
Cursor’s pricing strategy is also noteworthy, with the new model priced at $0.50 per million input/output tokens, compared to Composer 1.5’s $3.50 per million input tokens. A faster variant, Composer 2 Fast, costs $1.50 per million input/output tokens.
While Composer 2 may not be the top-scoring model in every benchmark, its strategic importance lies in its integration with Cursor’s tool stack and lower pricing. The company is positioning itself as a managed application layer that offers team features, governance, and workflow tooling on top of multiple model providers.
However, this move also puts pressure on Cursor to prove its value proposition, as first-party AI companies like OpenAI and Anthropic are increasingly shipping their own coding interfaces and agents. Some users have already expressed frustration with Cursor’s pricing and context loss, and are turning to alternative solutions like Anthropic’s Claude Code.
Despite these challenges, Composer 2 represents a significant step forward for Cursor, demonstrating its commitment to improving its in-house models and providing a more efficient cost-to-intelligence tradeoff for everyday coding work. As the AI landscape continues to evolve, Cursor’s strategic importance will depend on how well it can balance its focus on integration and value proposition with the needs of its users.
Source: https://venturebeat.com/technology/cursors-new-coding-model-composer-2-is-here-it-beats-claude-opus-4-6-but