Python 3.14: Free threaded Python is here (blog.python.org)

Python 3.14.0 is released, introducing a slate of features that change how CPython handles concurrency, performance and tooling. The headline is PEP 779 — “free-threaded” Python — now officially supported, alongside PEP 734 (multiple interpreters in the stdlib) and a new optional interpreter type that can deliver notably better performance on modern compilers. Official macOS and Windows binaries also ship an experimental JIT. Other important additions include deferred evaluation of annotations (PEP 649), template string literals (PEP 750), a stdlib Zstandard compression module (PEP 784), a zero‑overhead external debugger interface (PEP 768), improved error messages, and improved HMAC with formally verified code from HACL*. For AI/ML practitioners, the release is significant because free-threading and multiple interpreters reduce the need for heavyweight multiprocessing workarounds to get parallel CPU-bound throughput, simplifying data pipelines and server concurrency (though the new interpreter and modes are opt‑in and may require building from source). The experimental JIT and new interpreter path may improve numerical and model-serving performance over time. Added tooling — remote pdb attach, async process inspection, robust compression (zstd) for artifacts, and deferred annotations for cleaner typing — make development, deployment and debugging of ML services more efficient. Note build-chain changes (Sigstore recommended for release verification) and some breaking deprecations; review the “What’s new in Python 3.14” notes before upgrading.