cursor composer 2 model is a lot cheaper than Claude Opus
"tokens are the currency of LLMs", and as usage increases so do costs, and these costs are no longer small change for organizations, t he messaging has definitely shifted to being cost conscious. cursors Composer 2 model is cursors own model available in cursor cursors composer model is cheap, 10x cheaper than Claude Opus 4.6/4.7 per million output tokens $2.5 versus $25 (and is also 10x cheaper for input tokens) 6x cheaper than Sonnet wow! that's really significant and from the chart below, from cursor benchmarks, indications are it performs well compared to Opus 4.6 imo it makes total sense to use cursors Composer 2 model as default why not "Auto"? I've not used Auto so I can't comment from experience but I have read that it can be unpredictable and costly. Composer 2 works perfectly well for me If in Claude code then use Opus for complex tasks otherwise user Sonnet e.g. "write unit tests for Component Y" use Sonnet