NEW YORK — What happened: Python 3.15 has arrived with the experimental free-threaded build mode now fully optimized and widely supported by mainstream scientific and web frameworks. Major Linux distributions and package managers are beginning to default to the GIL-free interpreter for performance-critical environments.
Why it matters: Historically, Python's Global Interpreter Lock prevented multi-threaded CPU-bound code from utilizing multi-core processors efficiently. The removal of the GIL unlocks true parallel execution for data science, machine learning preprocessing, and concurrent web servers without requiring complex multi-process setups.
Deep dive: Moving away from the GIL requires C-extension modules to explicitly declare thread safety. In Python 3.15, core libraries like NumPy, Pandas, and Django have adopted thread-safe reference counting, preventing the segmentation faults that plagued early trial versions in previous releases.
Report check (claims vs what is verified vs still rumor): While core developers claim single-threaded performance overhead is now under two percent, independent benchmarks show a minor five to eight percent slowdown on pure single-threaded integer math due to atomic operations. Claims that legacy C extensions will work automatically without recompilation are false.
Open questions: Will third-party library maintainers drop support for the legacy GIL-enabled runtime sooner than expected, and how will memory consumption patterns shift under heavy multi-threaded workloads?