For DNA? Not sure. https://pmc.ncbi.nlm.nih.gov/articles/PMC3197634/ claims it's reached parity on protein sequences.

I'd much rather have a slower, but more accurate searcher, or one that was easier to use as an API.

There have been countless specializations and improvements on the original BLAST (which itself evolved).

When I left the field, the latest hottest thing was diamond (https://github.com/bbuchfink/diamond).