Python 3.15’s JIT gains stand out in a small benchmark

Miguel Grinberg compares Python 3.15’s third release candidate with earlier interpreters using recursive Fibonacci and bubble-sort programs. In his tests, the ordinary interpreter changes little from 3.14, while the experimental just-in-time compiler improves more clearly. Free-threaded builds help the parallel CPU-bound versions, but the benefit differs substantially between the two algorithms.

Miguel Grinberg · Oct 5, 2026

Why it matters

The results show why interpreter choice and workload matter more than a single headline speed number.

More context

The tests ran on one Intel Core i5 Gentoo Linux laptop, averaging three runs, with single-threaded and four-thread variants. Try comparable measurements on your own application before drawing upgrade conclusions; the author’s programs are useful for studying the method.

Keep in mind

These are two informal microbenchmarks, not a broad application suite, and exclude I/O-bound work. The tested interpreter is a release candidate and the JIT remains experimental. We read the results rather than rerunning them.

Original read