python training

The latest Python version is 3.15, released on 9 October 2026. Here are its biggest changes compared with 3.14:

1. Explicit lazy imports (PEP 810)

The new lazy soft keyword delays loading a module until the imported name is first used. This can improve startup time, while keeping imports at the top of the file.

lazy import json
lazy from pathlib import Path

print("Starting...")          # Neither module is fully loaded yet
data = json.loads('{"a": 1}') # json loads here

lazy works only at module level, so using it inside a function, class, or try block raises a SyntaxError. Star imports can't be lazy either. If a lazy import fails, the error appears at first use rather than at the import line.

2. Built-in frozendict (PEP 814)

frozendict is an immutable dictionary. It is hashable if all its keys and values are hashable, and it isn't a subclass of dict. It keeps insertion order, but order is ignored when comparing two objects.

config = frozendict(host="localhost", port=8080)
cache = {config: "connected"}  # Can be used as a dict key
config["port"] = 9000          # TypeError

3. Built-in sentinel (PEP 661)

The new sentinel type creates unique marker values that have a concise representation and keep their identity when copied. It's useful when None is a valid value.

MISSING = sentinel("MISSING")

def get_value(key, default=MISSING):
    ...

4. Unpacking in comprehensions (PEP 798)

Comprehensions and generator expressions now support * and ** unpacking.

lists = [[1, 2], [3, 4], [5]]
flat = [*items for items in lists]    # [1, 2, 3, 4, 5]

dicts = [{"a": 1}, {"b": 2}, {"a": 3}]
merged = {**d for d in dicts}         # {"a": 3, "b": 2}

5. UTF-8 by default (PEP 686)

When you don't specify an encoding, operations like open("file.txt") now use UTF-8, regardless of the system environment. This matters most for Windows code that relied on the old locale-based default. To restore the old behavior, use PYTHONUTF8=0 or -X utf8=0. Passing encoding= explicitly is still the safest choice across versions.

6. New profiling tools, including Tachyon (PEP 799)

The new profiling package includes profiling.tracing for deterministic tracing and profiling.sampling for statistical sampling.

The sampling profiler, Tachyon, can attach to a running process without code changes or a restart, which makes it useful for production debugging. It supports flame graphs, heatmaps, a live terminal view, and async-aware profiling.

7. Faster experimental JIT

The experimental JIT compiler was significantly upgraded. Reported gains are about 7–8% on x86-64 Linux and 11–12% on AArch64 macOS.

8. Better error messages

Python now suggests fixes for more mistakes, including method names from languages like JavaScript, Java, Ruby, and C#.

[1, 2, 3].push(4)
# AttributeError: 'list' object has no attribute 'push'. Did you mean '.append'?

9. Other notable changes

• Typing: Additions include TypedDict with typed extra items (PEP 728), TypeForm (PEP 747), and disjoint bases (PEP 800).
• Free-threading: C extensions can now target a Stable ABI for free-threaded builds, known as abi3t (PEP 803).
• Async: TaskGroup.cancel lets you stop a task group early.
• Observability: Frame pointers are on by default where supported, improving native profiling and debugging (PEP 831).
• Tooling: More command-line tools now use color, and the interactive shell has better tab completion.

Before you upgrade

• The profile module is deprecated and scheduled for removal in Python 3.17, while cProfile remains as an alias.
• re.match() is soft deprecated in favor of the clearer re.prefixmatch(), but soft deprecation doesn't mean it's being removed.
• Check the official porting guide, especially if your code relies on the old default encoding.

Sources: What's new in Python 3.15 – Python documentation | Python 3.15.0 release report – Zeli