Source: LWN.net systems report titled 'Python's experimental JIT to be removed' published October 27, 2026.

First introduced behind an opt-in compiler flag in Python 3.13, the copy-and-patch JIT was developed by the Faster CPython team to provide lightweight native machine code generation. By parsing precompiled C object files at build time and emitting pre-assembled machine instructions at runtime, the system sought to deliver JIT acceleration without requiring CPython binaries to bundle heavyweight LLVM or GCC dependencies.

However, integrating the copy-and-patch architecture into CPython's production build pipeline created friction across the developer ecosystem. In July 2026, core contributors Brandt Bucher, Haoyi Ostrowski, and Ken Jin published PEP 836 to formalize the JIT's operational governance. The proposal sparked deep concerns from Linux distribution packagers and cross-compilation maintainers regarding specialized toolchain requirements and increased binary footprints.

Crucially, empirical performance benchmarks failed to justify the engineering maintenance overhead. Across typical web services, data manipulation, and async networking benchmarks, the copy-and-patch JIT provided only modest 2% to 7% execution speed improvements, while sometimes degrading memory usage and warmup latencies. In contrast, CPython's Tier 1 and Tier 2 specializing adaptive interpreter yielded 20% to 30% performance gains purely through bytecode optimization.

Following an extensive community review cycle, the Python Steering Council determined that maintaining the copy-and-patch JIT alongside emerging Tier 2 optimizations was counterproductive to long-term language health, directing core developers to remove the JIT code prior to the Python 3.15 release.

The Faster CPython project will reallocate engineering resources toward trace-based execution models. By optimizing linear execution traces directly within Tier 2 micro-ops, Python developers aim to achieve robust execution speedups while maintaining cross-platform portability across embedded and enterprise systems.