2x or 5x is very useful if it speeds up common web server applications. A company with 4 servers can run 2 servers with a 2x speedup. Math-heavy benchmarks are

2x or 5x is very useful if it speeds up common web server applications. A company with 4 servers can run 2 servers with a 2x speedup. Math-heavy benchmarks are not really representative of what Python is used for. They benefit from things like SIMD and aggressive inlining and fine-grained control over memory layout, etc. If you need that with Python, you can use something like NumPy or implement the hot path with a native language.