[_factor]

It’s in interesting question. I assume they’re using accelerators, and the alternative is a disk or memory hit. It still seems expensive to me.

https://www.linkedin.com/pulse/rising-cost-llm-based-search-...