How Fast is Python 3.15?

This is an informal performance benchmark that compares Python 3.15 (release candidate) against CPython 3.10-3.14, PyPy 3.12, Node.js 26.3 and Rust 1.97 using two test programs: a recursive Fibonacci computation (fibo.py, first 40 terms) and a bubble sort of 10,000 random numbers (bubble.py). Each test is run single-threaded and with four parallel threads, and CPython variants include standard, JIT (available since 3.13) and free-threading (no GIL). Tests were run three times on an Intel Core i5 laptop running Gentoo Linux and results reported as average times and relative speed ratios versus Python 3.15.

Findings: CPython 3.15 shows only a small single-threaded improvement over 3.14; the most notable interpreter advances came in 3.11 and 3.14. The JIT build in 3.15 produces meaningful wins (about 1.20× faster on fibo, 1.28× on bubble versus standard 3.15). The free-threading build delivers large multithreaded gains (about 4.5× faster than standard CPython on the multi-threaded fibo test) by removing the GIL, though its single-threaded performance is similar to previous releases. PyPy remains dramatically faster than CPython on these workloads (e.g., ~5.5× faster on fibo), while Node and especially Rust outpace the Python runtimes by large margins on these specific tests.