Machine Learning and Visualization in Julia

The first chart has a graph of power, and rates Python as the least powerful among C/C++, Java, MATLAB, Go, Erlang, and Julia, besting only Mathematica.

I don't know how you define "power", and the article doesn't explain the methodology, but regardless of how you measure it, I would strongly argue that unless you are discounting any contribution from available libraries, the Python numerical processing ecosystem makes it the strongest of all of these. You state Cython is another language, and appear to give no credit to numba, which can be invoked in many cases for huge speedups with a simple decorator. You can't evaluate Python, or any language for that matter, without looking at the capabilities of its ecosystem.