Python's JIT Compiler Is Graduating: PEP 836 Charts Path to Official Support

Python core developers have published PEP 836, a roadmap to promote CPython's experimental JIT compiler from opt-in experiment to fully supported feature, targeting 20% speedup by Python 3.17.

The experimental Just-in-Time compiler that has been lurking inside CPython since version 3.13 is finally getting a real shot at graduation.

Python core developers Savannah Ostrowski, Ken Jin, and Brandt Bucher published PEP 836 on Wednesday, laying out a 2.5-year roadmap to transform the JIT from an opt-in experiment into a supported, non-experimental feature of the language. The PEP carries a characteristically understated title: "JIT Go Brrr."

The JIT has been quietly maturing. In Python 3.15, it already delivers a measurable 4–12% geometric mean performance improvement across Tier 1 platforms, reduces generated code's memory footprint compared to 3.14, and produces stack frames that native debuggers can unwind. But the team wants more — the PEP sets a minimum bar of 20% geometric mean improvement on pyperformance benchmarks by the first beta of Python 3.17, measured across all supported Tier 1 platforms.

The roadmap spans roughly two and a half years. Year one — targeting Python 3.16's first beta — focuses on developer experience: evolving the JIT's frontend from trace-based recording to method-based compilation, making it compatible with Python's free-threading (no-GIL) builds, and providing redistributors with a reproducible way to build JIT stencils without locking into a single LLVM version. The team also wants at least a 5% performance uplift for the JIT-plus-GIL build over the interpreter alone as an early checkpoint.

If the milestones are met, the JIT can be promoted to non-experimental status. If not, the Steering Council and core team will re-evaluate whether the JIT should remain in CPython's main branch at all.

This is a measured, milestone-gated approach — no blank checks. The PEP explicitly rejects several alternatives, including maintaining the JIT outside of CPython main or building a pluggable JIT infrastructure. The message is clear: the JIT either proves its worth inside the tree, on a fixed timeline, or it gets cut.

For a language that has spent three decades relying on a straightforward bytecode interpreter, this is a consequential moment. Python has absorbed lessons from the JavaScript world, where JIT compilers turned a slow scripting language into a performance contender. If the team hits its targets, Python 3.17 could ship with the biggest single-threaded performance leap in the language's history.

The PEP is currently in Draft status with discussions pending. The source is available on GitHub.