Python’s GIL Set to Disappear: What the No‑GIL Roadmap Means for Developers

Python will no longer be the pseudo‑multithreading language many have described.

"The GIL in Python will be gone – a huge win for the AI ecosystem," said PyTorch core maintainer Dmytro Dzhulgakov.

The Global Interpreter Lock (GIL) is a mutex that protects Python objects, preventing multiple threads from executing Python bytecode simultaneously, ensuring thread safety but limiting execution to one thread per CPU.

Its main drawback is that only one thread can run on a CPU at a time, preventing true multithreaded concurrency and reducing performance.

The Python core team has now accepted the proposal to make the GIL optional, a development welcomed by many developers.

Meta engineer Sam Gross spent over four years implementing the no‑GIL build.

Yann LeCun, one of the deep‑learning pioneers, celebrated the news, noting that without the GIL, Python code can execute multithreaded freely.

Thomas Wouters, a CPython core developer, outlined the details and future outlook of the no‑GIL proposal.

The basic plan includes:

In the long term (about five years or more), the no‑GIL build should become the sole build.

Careful handling of backward compatibility, ensuring third‑party code changes for the no‑GIL build only affect the with‑GIL build, while still addressing older Python versions.

Community support is required before fully committing to the no‑GIL default; the core team needs experience with the new build mode, thread safety, and the new C and Python APIs.

If the no‑GIL mode proves too disruptive, the team may revert, keeping no‑GIL specific code identifiable.

The roadmap is divided into three stages:

Short term: an experimental no‑GIL build expected in Python 3.13 (or possibly 3.14), not set as the default.

Mid term: after sufficient community support, the no‑GIL build will be supported but not default, with a target release date to be decided.

Long term: aim for the no‑GIL build to become the default, removing all traces of the GIL while maintaining necessary compatibility, a process projected to take about five years.

The development team will continuously evaluate progress and adjust timelines as needed.

What are your thoughts on making the GIL optional?