Lies, damn lies and benchmarks - I'm not saying the benchmarks are made up but you know - the benchmark thing.
Falcon for REST API development. Django/Pyramid for full featured web application development.
And regarding speed, well how does its Python 3 performance compare to the Python 2 performance of other Python web servers? And any does speed matter that much? In many cases "fast enough is good enough" and the bottle necks are more likely to be other parts of your system and if speed matter that much then you should be horizontally scaling.
And if I may raise the ugly topic - who cares about Python 2 under any circumstances? For me - and I know many people do not think this way - Python 2 is dead so who gives a rats arse anything about it. Python 3 is the future and I'm not looking back for any gripes or grumbles about it.
Don't get me wrong, I think ruby-rails-slow is a bad thing, but Falcon fast should be good enough.
And if speed really matters to you then you shouldn't be using Python for web application development.
In all my research over the years Pyramid is the framework that looks the most interesting. The docs are amazing (I more or less read them cover to cover, even though I haven't tried the framework) and the whole thing seems very well thought out.
I'd love any insights from people who have worked with it.
Falcon gives you less than Flask, not more.
For me the big thing is that Falcon has none of the stuff needed for building web apps. Flask has all that stuff cause it was built when back end servers also built the HTML pages. There's alot of stuff associated with that functionality in Flask that is not there in Falcon. It makes it much cleaner.
Personally, I'm excited to explore flask more, since it seems to put the ball entirely in your court. I like the concept behind flask a lot -- for some projects I'm willing to sacrifice additional framework features in exchange for a more careful understanding of everything I am adding to my app.
The problem as I see it is that Pyramid has a lot of extra interfaces and support for things in the core by default (and some concepts that are completely missing from most frameworks like context). So much is there that it's easy to lose sight of how little is actually required. Fundamentally you can write Pyramid apps in 10 lines of code exactly as you would with Flask but that's less interesting for bigger apps and so the docs focus on all the extra features.
I mean if you take the example at the top of http://docs.pylonsproject.org/projects/pyramid/en/latest/ and add in the pyramid_jinja2 library via config.include('pyramid_jinja2') then you basically have views and jinja2 rendering fully baked.
As in the previous example if you get right down to it Pyramid doesn't even have a templating engine! Flask comes with jinja2 built-in! So scaling DOWN to APIs where this is not necessary is an interesting thing..
By saying "it's not a particularly opinionated framework", I mean to say that it allows you to add, easily, whatever you want to it. However, there is a fair amount of conceptual load in doing so. The difference between a flask app and a pyramid app with pyramid jinja2 built in is that one of them takes about 20 seconds to understand.
Don't get me wrong -- I think pyramid is fantastic. But it it is disingenuous to suggest that there are not differences in complexity between the two frameworks.
I'm interested in the present.
You of course may use whichever software you want but I don't have enough time to learn anything but what is currently being developed.
I'd take those benchmarks with a massive pinch of salt. 338µs per request for Flask is the bad case. 0.5ms of the request is spent in the framework, that's not even noise. The fastest framework I've used is one I wrote for coldfusion and that's the only time I've expected all my requests to be sub 100ms. And even then, the network latency turns that into noise.
Look at features before performance - both in the case of frameworks and version of Python to use (Python 3 is a lot nicer).