Python 3.15’s JIT hits speed targets early on Apple Silicon

Python 3.15’s just-in-time compiler is already outpacing expectations, delivering on its modest performance goals for macOS AArch64 a full year early and for x86_64 Linux several months ahead of schedule. The alpha JIT, according to core developer Ken Jin, clocks in at 11–12% faster than the tail-calling interpreter on Apple Silicon and 5–6% faster than the standard interpreter on x86_64 Linux.

Python’s interpreter has long been its Achilles’ heel, but these gains suggest the language’s new JIT is finally closing the gap. The CPython team’s achievable targets hint at steady progress rather than sudden leaps—a relief for developers weary of hype cycles around performance claims. Benchmarks like these matter because they translate to real-world gains in latency-sensitive workloads, from web servers to data pipelines.

The modest goals, as Jin frames them, underscore a deliberate approach to optimization without overpromising. It’s a pragmatic shift from the days when language updates promised revolutions in speed with little to show for it.

What’s striking here is how the JIT’s gains align with broader industry trends toward heterogeneous computing. Apple’s AArch64 chips have forced optimizations that often benefit other platforms, while x86_64 systems gain incrementally but consistently. For organizations running Python at scale, even a 5% speedup can mean shaving minutes off long-running jobs—critical when costs scale with CPU cycles.

The JIT’s performance ceiling remains an open question, but these early results suggest it’s raising the floor. Python’s interpreter has been the primary bottleneck since its inception, and incremental gains here could redefine expectations for what’s possible in everyday workloads. The real signal isn’t the size of the speedup, but the consistency of the progress.

The bigger question is whether this momentum will sustain through Python 3.15’s full release. Will the gains hold under varied workloads, or will edge cases expose hidden trade-offs? The community is watching closely.