It's unclear why the techempower gives low performance for elixir / phoenix. But from what I've heard is that in real world applications the performance tends to scale much better than Python or Ruby.

it's because the people who contribute to techempower benchmarks are zealots who believe that it's far more useful and important to show their One True Language in the Best Possible Light than to give you a useful benchmark. take a look at the source code for some of them and see for yourself.

The link is about execution speed, I think if you start talking about 'scale' that's a diversion, since that isn't about the language and is more about the architecture of the program.

These benchmarks have been thoroughly studied by the elixir community (even myself) and are... Trash.

They compare things that have nothing to do with production, use non realistic code for most language and keep refusing PR to fix it for elixir.

The community simply decided to stop trying to engage and let that code die.