Turns out there are some optimizations you can do when compiling Python 3.5 and 3.6 that give you some significant speed improvements without any real downside.
We've paired with with Google's base Debian image that rips out systemd and all it's dependencies, which you weren't going to be using in a Docker container anyway, which yields a MUCH smaller image.
How do you use it? Pretty simple just start your Dockerfile with:
And rebuild your image.
How much faster is it? Depending on the benchmark in question it's up to 19% faster and nearly 75% smaller than the official Python images.