Introducing Cursor 2.0. Our first coding model and the best way to code with agents.
Composer is a frontier coding model that completes tasks in under 30 seconds.
Easily run the same prompt across models for maximum intelligence.
With a built-in browser, agents can now run and test their code.
Turn speech into code with voice mode.
Replying to @cursor_ai
That was a fast jump from 1.0! When’s 3.0? cursor.com/changelog/1-0
2
Replying to @cursor_ai
Absolutely insane - can’t wait to start using it, the multi-model functionality is something I was really hoping was coming. Safe to say this will likely be the highlight of my next issue of AIPoweredEngineering.com
1
Replying to @cursor_ai
Great - how’s the benchmarks for composer?
1
Replying to @cursor_ai
Will my cursor update to this or is this a separate app?
1
Replying to @cursor_ai
@geekslyy case in point here
And to add, I don't think you should start with building an AI powered app with an SLM. Begin with an LLM to get a sense of performance and then distill into smaller components when you've gotten a good sense of model performance as its more of a science
2
1
Replying to @cursor_ai
this is amaaaaazinggg
Replying to @cursor_ai
Congrats on the launch. I'm most interested in the browser integration, but it appears completely disjointed. Open the browser tab, and the agent simply ignores it, attempts to open it's own browser and then proceeds on its own. We really need full browser integration so human/agent are fully interchangable.
Replying to @cursor_ai
Woah a built in browser is a game changer!
Replying to @cursor_ai
Woah that's a lot of changes, crazy work
Replying to @cursor_ai
How much is the new cursor model to use ?
Replying to @cursor_ai
Great release!
Replying to @cursor_ai
Curious to see what changes made it in. Anything you're most excited about for tomorrow?
Replying to @cursor_ai
Voice mode doesn’t work
Replying to @cursor_ai
any plans to support multi agentic workflows with agents interacting?
Replying to @cursor_ai
Please hosting it on Cerebras, it would be wild to have model run at 2000 tk/s