Top New Features in Python 3.15: Lazy Imports, Faster JIT, Better Errors, Smarter Profiling

Python 3.15’s first full release is one of the most feature‑rich versions of the language, bringing a suite of long‑awaited enhancements.

Lazy Import

The new lazy‑import syntax lets modules be imported only when their code is actually executed, reducing start‑up time for heavy imports. Developers can use the explicit lazy‑import statement or enable it via an environment variable, and the behavior matches normal imports without drawbacks.

Built‑in frozendict

A new immutable, hashable dictionary type called frozendict behaves like a regular dict but cannot be modified, allowing it to be used safely as a key in other dictionaries.

Built‑in sentinel() Type

Python now provides a sentinel() factory that creates unique objects (e.g., None, object(), sentinel("NAME")) to replace the common pattern of using is with a unique sentinel. These objects have informative representations and support proper type checking.

Statistical Sampling Profiler

The new profiling.sampling module offers statistical sampling profiling with minimal runtime impact, complementing the existing deterministic cProfile (still available as profiling.tracing).

Upgraded JIT

The JIT compiler, first introduced in Python 3.13, now delivers a geometric‑mean speed improvement of roughly 8%‑13% over standard CPython, depending on platform and workload. Gains stem from a new tracing frontend, register allocation, higher‑quality machine code, and optimizations such as eliminating ref‑count updates for certain object classes.

Clearer Error Messages

Error messages have become more precise. For missing attributes, the interpreter suggests possible correct names (e.g., "Did you mean 'xyz'?") and also checks for deleted attributes. When fuzzy name matching fails, it falls back to a list of common method names from other languages, suggesting .append() instead of the erroneous list.push().

Type System Improvements

TypedDict

now accepts closed (restrict runtime keys) and extra_items (declare additional keys with explicit types). The new TypeForm construct lets type expressions be evaluated to values, enabling uses such as typing.cast or isinstance in places where a type is treated as a value.

List‑Comprehension Unpacking

Flattening nested iterables can now be written with the star operator inside a comprehension, e.g., y = [*a for a in x]. The double‑star ** operator works for dictionary merging: y = {**d for d in dicts}. The same unpacking syntax can build generator expressions.

Garbage‑Collection Rollback

Python 3.14 introduced an incremental garbage collector that reduced pause times but increased overall memory usage for many users. Python 3.15 reverts to the classic generational collector, with the incremental collector slated to return only after further refinements.