CHALLENGE 2: Benchmark Cursor’s Models

Level: Beginner

Use Cursor’s different models and compare their performance. Always check cursor.com/docs for the latest.

What are we building?

This challenge is fairly open-ended. You can build anything you’d like, but the idea is to stay consistent. Choose a single prompt and run it up to 3 times in Cursor, using a different model for each attempt.

How it works

1. Go to cursor.com and click Download (it auto-detects your OS).
2. Open the AI chat — it lives right in the editor.
3. Toggle the model you want to test at the bottom of the chat window.
4. Watch it respond — with code or an explanation, in context of your project.
5. Note the results.

Steps

1. Open the Cursor Desktop App and select New Chat.
2. In the dropdown menu above the chat window, select New Folder to create a new project folder.
   • Optional: Open the IDE and open the side chat for a new agent.
3. In the chat window, select the Model you want to test.
4. Your prompt can be something like the following:
   • Paste in a chunk of code and ask “explain what this does.”
   • Ask “write a function that reverses a string” (in any language you like).
5. Note the results and repeat steps 1 through 4 with the same prompt but a different model.

What to Submit

• A description of the prompts you ran including the models you chose (e.g., Grok 4.7, Composer 2.5, etc.) and the pros and cons for each one.
• A screenshot of your results.