Python 3.15 JIT beat the standard interpreter in Grinberg's benchmarks

The JIT in Python 3.15 ran faster than the standard interpreter in Miguel Grinberg's benchmarks, a result that was not true in any previous release. Grinberg has run the same two tests, recursive Fibonacci and bubble sort, each year since Python 3.10, across the standard and JIT variants plus PyPy, Node.js and Rust, and the 3.15 numbers are the first where turning the JIT on helps rather than hurts.

The post is a direct side-by-side comparison, and Grinberg is explicit about what he measured and what he did not. The result worth reading is the one comparing Python 3.15 against earlier CPython releases, which is where this year's finding sits.

What was measured, and how

Grinberg ran two small programs: a recursive Fibonacci calculation and a bubble sort. Both ran in single-threaded and multi-threaded variants. The runtimes under test were CPython 3.10 through 3.15, including standard, JIT and free-threading builds where they existed; PyPy 3.12; Node.js 26.3; and Rust 1.97. Each test ran three times and the averages were reported as ratios against a baseline, with 1x being the baseline speed.

The headline finding is that Python 3.15 with the JIT turned on ran roughly 1.20 to 1.28x faster than Python 3.15 with the standard interpreter. In previous years, the JIT either did not exist or ran slower than the standard build in Grinberg's tests. A gap of 1.20 to 1.28x on a microbenchmark is not production throughput, but it is the first time switching the JIT on makes a Python program faster rather than slower in this test, which is the news.

What Grinberg argues a reader should actually do

Grinberg's advice is to not rush. The standard 3.15 interpreter shows only a small gain over 3.14, so if you are upgrading purely for speed, the case is thin. He suggests waiting until the JIT exits experimental status, or until Python 3.16 is released, before migrating a production service. If you want the 3.15 language features, upgrade and use them. If you want a free throughput gain on CPU-bound Python, the JIT variant on 3.15 is now worth adding to a profiling run.

PyPy and Rust still beat both standard and JIT CPython by large multiples on both benchmarks. The real audience for Grinberg's result is a team already on CPython that cannot easily move to another runtime: the JIT build is the first option on 3.15 that produces a measurable throughput gain over the standard interpreter.

Source

Primary: How Fast is Python 3.15?, Miguel Grinberg.