Sounds like wishful thinking. I wish it was true too, but it isn’t. Or is there one reliable source which has valid benchmarks showing that they are similar? Mo

Sounds like wishful thinking. I wish it was true too, but it isn’t. Or is there one reliable source which has valid benchmarks showing that they are similar? Most benchmarks I’ve seen appear cherry-picked (as is the case in ML for years already. Be skeptical of benchmarks).