I must confess -- in my Data Science oriented mind, I am not often thinking about things like energy efficiency when it comes to these things. That certainly is an interesting metric, but I certainly don't have the answers -- at least not for these frameworks in particular.
After some prying around, I was able to find a pretty concise look at energy efficiency across different languages, including Julia here:
https://sites.google.com/view/energy-efficiency-languages/updated-functional-results-2020?pli=1
CPU usage seems to remain low with all of them. I know that's not the best metric for efficiency, but I'm sure it can give some level of inclination. Toolips tends to use less than 0.2 percent of my tiny server processor with an RPC chat application -- so I don't think that one would be too bad.
I think if you are comparing it to something like Python, it is certainly going to be more efficient. However, of course there are always more efficient solutions out there -- just a lot more work to build such things with them.
I am sorry I don't exactly have an answer to this question, I will have to find some way to benchmark such things and perform a test perhaps, it certainly would be interesting. Thank you for your thoughtful and provoking comment, it certainly wasn't a metric I had thought of before in these considerations -- and is certainly an important one. I appreciate it!