Cursor launched Composer 2 in March 2026. It is Cursor's own frontier-level agentic model, built to rival OpenAI and Anthropic for coding tasks. If you use Cursor for AI-assisted development, here is what changed and why it matters.
What is Cursor Composer 2?
Composer 2 is Cursor's third-generation agentic coding model. It runs inside the Cursor IDE and handles tasks that go beyond simple code completion: reasoning through problems, editing multiple files, running terminal commands, and maintaining context across long sessions. The "2" marks a major step up in quality and efficiency.
The model is tuned for tool use, file edits, and terminal operations. It is trained on long-horizon coding tasks through reinforcement learning and self-summarization, so it can solve challenging tasks that require hundreds of actions without losing track.
Benchmark improvements
Cursor publishes results on CursorBench, Terminal-Bench 2.0, and SWE-bench Multilingual. Composer 2 shows large gains over Composer 1.5 and the original Composer:
| Model | CursorBench | Terminal-Bench 2.0 | SWE-bench Multilingual |
|---|---|---|---|
| Composer 2 | 61.3 | 61.7 | 73.7 |
| Composer 1.5 | 44.2 | 47.9 | 65.9 |
| Composer 1 | 38.0 | 40.0 | 56.9 |
These improvements come from continued pretraining and reinforcement learning on long-horizon tasks. Composer 2 is able to handle complex, multi-step work that earlier versions struggled with.
Pricing and tiers
Composer 2 is priced at $0.50/M input and $2.50/M output tokens. Cursor also offers a faster variant with the same intelligence at $1.50/M input and $7.50/M output tokens. The fast variant is the default in the product and is priced lower than other fast models at similar speeds.
On individual plans, Composer usage is part of a standalone usage pool with generous included usage. On team and enterprise plans, the API price is charged directly. Full details are in the Composer 2 model docs.
Key features for developers
Composer 2 keeps the strengths of Composer 1.5 and adds more:
- Tool use: Full access to agent tools in Cursor, including file edits, terminal commands, and codebase search.
- Long-horizon tasks: Self-summarization during training helps the model stay on task across hundreds of actions.
- Two speed tiers: Standard tier for cost per token; fast tier for interactive sessions.
- Same workflows: You can use the same prompts and workflows as with Composer 1.5. The improvements are in how the model reasons and executes.
For practical habits that work with any AI assistant, see our AI-assisted coding tips.
How to try Composer 2
Composer 2 is available in Cursor today. To use it:
- Update Cursor: Make sure you are on the latest version.
- Select the model: In Composer, choose Composer 2 (or Composer 2 Fast) from the model selector.
- Optional: Try the early alpha of Cursor's new interface for a different experience.
No extra configuration is required. If you already use Composer, you can switch to Composer 2 and continue with your existing workflows.
When Composer 2 makes sense
Composer 2 fits developers who:
- Work on multi-file refactors, bug hunts, or feature builds
- Need reliable terminal and tool use inside the IDE
- Want frontier-level coding intelligence without switching to a different tool
It is overkill for very simple edits or when you already have a workflow that works well. For most day-to-day coding with AI assistance, Composer 2 offers a strong balance of speed, quality, and cost.
Cursor Composer 2 is a clear step up from 1.5. If you rely on Cursor for coding, it is worth trying the new model and seeing how it handles your real tasks.