Hello,
This chart was published with the announcement about composer2. I think it’s very useful and it’d be great if we could have the same for all the models available in cursor.
Is there a place where we can find such a thing ?
1 Like
Hey there!
You can read more about CursorBench, and see results for a lot more models, in this blog post:
1 Like
Cursor, please consider publishing results of CursorBench for more models, more often. For example I’m very curious how gpt-5.3-codex-spark performs on your bench. I’ve tried it, and it felt much less pleasant to use compared to Composer 2 Fast. I’d like to see if CursorBench aligns with my feelings.
Also, I’m working a lot with agentic coding, and the bottleneck for my workflow is reviewing code the models generate, so I cannot invest a lot of my time in trying out different models. If the results of CursorBench were published regularly for more models, it’d be easier for me to simply pick the best model for my work, and save my time.
1 Like