What’s new in Python 3.14

Editors:

Adam Turner and Hugo van Kemenade

This article explains the new features in Python 3.14, compared to 3.13.Python 3.14 was released on 7 October 2025.For full details, see thechangelog.

See also

PEP 745 – Python 3.14 release schedule

Summary – Release highlights

Python 3.14 is the latest stable release of the Python programminglanguage, with a mix of changes to the language, the implementation,and the standard library.The biggest changes includetemplate string literals,deferred evaluation of annotations,and support forsubinterpreters inthe standard library.

The library changes include significantly improved capabilities forintrospection in asyncio,support for Zstandard via a newcompression.zstd module, syntax highlighting in the REPL,as well as the usual deprecations and removals,and improvements in user-friendliness and correctness.

This article doesn’t attempt to provide a complete specificationof all new features, but instead gives a convenient overview.For full details refer to the documentation,such as theLibrary ReferenceandLanguage Reference.To understand the complete implementation and design rationale for a change,refer to the PEP for a particular new feature;but note that PEPs usually are not kept up-to-dateonce a feature has been fully implemented.SeePorting to Python 3.14 for guidance on upgrading fromearlier versions of Python.


Interpreter improvements:

Significant improvements in the standard library:

C API improvements:

Platform support:

Release changes:

New features

PEP 649 &PEP 749: Deferred evaluation of annotations

Theannotations on functions, classes, and modules are nolonger evaluated eagerly. Instead, annotations are stored in special-purposeannotate functions and evaluated only whennecessary (except iffrom__future__importannotations is used).

This change is designed to improve performance and usability of annotationsin Python in most circumstances. The runtime cost for defining annotations isminimized, but it remains possible to introspect annotations at runtime.It is no longer necessary to enclose annotations in strings if theycontain forward references.

The newannotationlib module provides tools for inspecting deferredannotations. Annotations may be evaluated in theVALUEformat (which evaluates annotations to runtime values, similar to the behavior inearlier Python versions), theFORWARDREF format(which replaces undefined names with special markers), and theSTRING format (which returns annotations as strings).

This example shows how these formats behave:

>>>fromannotationlibimportget_annotations,Format>>>deffunc(arg:Undefined):...pass>>>get_annotations(func,format=Format.VALUE)Traceback (most recent call last):...NameError:name 'Undefined' is not defined>>>get_annotations(func,format=Format.FORWARDREF){'arg': ForwardRef('Undefined', owner=<function func at 0x...>)}>>>get_annotations(func,format=Format.STRING){'arg': 'Undefined'}

Theporting section contains guidanceon changes that may be needed due to these changes, though in the majority ofcases, code will continue working as-is.

(Contributed by Jelle Zijlstra inPEP 749 andgh-119180;PEP 649 was written by Larry Hastings.)

See also

PEP 649

Deferred Evaluation Of Annotations Using Descriptors

PEP 749

Implementing PEP 649

PEP 734: Multiple interpreters in the standard library

The CPython runtime supports running multiple copies of Python in thesame process simultaneously and has done so for over 20 years.Each of these separate copies is called an ‘interpreter’.However, the feature had been available only throughtheC-API.

That limitation is removed in Python 3.14,with the newconcurrent.interpreters module.

There are at least two notable reasons why using multiple interpretershas significant benefits:

  • they support a new (to Python), human-friendly concurrency model

  • true multi-core parallelism

For some use cases, concurrency in software improves efficiency andcan simplify design, at a high level.At the same time, implementing and maintaining all but the simplest concurrencyis often a struggle for the human brain.That especially applies to plain threads (for example,threading),where all memory is shared between all threads.

With multiple isolated interpreters, you can take advantage of a classof concurrency models, like Communicating Sequential Processes (CSP)or the actor model, that have foundsuccess in other programming languages, like Smalltalk, Erlang,Haskell, and Go. Think of multiple interpreters as threadsbut with opt-in sharing.

Regarding multi-core parallelism: as of Python 3.12, interpretersare now sufficiently isolated from one another to be used in parallel(seePEP 684). This unlocks a variety of CPU-intensive use casesfor Python that were limited by theGIL.

Using multiple interpreters is similar in many ways tomultiprocessing, in that they both provide isolated logical“processes” that can run in parallel, with no sharing by default.However, when using multiple interpreters, an application will usefewer system resources and will operate more efficiently (since itstays within the same process). Think of multiple interpreters ashaving the isolation of processes with the efficiency of threads.

While the feature has been around for decades, multiple interpretershave not been used widely, due to low awareness and the lack of astandard library module. Consequently, they currently have severalnotable limitations, which are expected to improve significantly nowthat the feature is going mainstream.

Current limitations:

  • starting each interpreter has not been optimized yet

  • each interpreter uses more memory than necessary(work continues on extensive internal sharing between interpreters)

  • there aren’t many optionsyet for truly sharing objects or otherdata between interpreters (other thanmemoryview)

  • many third-party extension modules on PyPI are not yet compatiblewith multiple interpreters(all standard library extension modulesare compatible)

  • the approach to writing applications that use multiple isolatedinterpreters is mostly unfamiliar to Python users, for now

The impact of these limitations will depend on future CPythonimprovements, how interpreters are used, and what the community solvesthrough PyPI packages. Depending on the use case, the limitations maynot have much impact, so try it out!

Furthermore, future CPython releases will reduce or eliminate overheadand provide utilities that are less appropriate on PyPI. In themeantime, most of the limitations can also be addressed throughextension modules, meaning PyPI packages can fill any gap for 3.14, andeven back to 3.12 where interpreters were finally properly isolated andstopped sharing theGIL. Likewise, libraries on PyPI are expectedto emerge for high-level abstractions on top of interpreters.

Regarding extension modules, work is in progress to update some PyPIprojects, as well as tools like Cython, pybind11, nanobind, and PyO3.The steps for isolating an extension module are found atIsolating Extension Modules.Isolating a module has a lot of overlap with what is required to supportfree-threading, so the ongoingwork in the community in that area will help accelerate supportfor multiple interpreters.

Also added in 3.14:concurrent.futures.InterpreterPoolExecutor.

(Contributed by Eric Snow ingh-134939.)

See also

PEP 734

PEP 750: Template string literals

Template strings are a new mechanism for custom string processing.They share the familiar syntax of f-strings but, unlike f-strings,return an object representing the static and interpolated parts ofthe string, instead of a simplestr.

To write a t-string, use a't' prefix instead of an'f':

>>>variety='Stilton'>>>template=t'Try some{variety} cheese!'>>>type(template)<class 'string.templatelib.Template'>

Template objects provide access to the staticand interpolated (in curly braces) parts of a stringbefore they are combined.Iterate overTemplate instances to access their parts in order:

>>>list(template)['Try some ', Interpolation('Stilton', 'variety', None, ''), ' cheese!']

It’s easy to write (or call) code to processTemplate instances.For example, here’s a function that renders static parts lowercase andInterpolation instances uppercase:

fromstring.templatelibimportInterpolationdeflower_upper(template):"""Render static parts lowercase and interpolations uppercase."""parts=[]forpartintemplate:ifisinstance(part,Interpolation):parts.append(str(part.value).upper())else:parts.append(part.lower())return''.join(parts)name='Wenslydale'template=t'Mister{name}'assertlower_upper(template)=='mister WENSLYDALE'

BecauseTemplate instances distinguish between static strings andinterpolations at runtime, they can be useful for sanitising user input.Writing ahtml() function that escapes user input in HTML is an exerciseleft to the reader!Template processing code can provide improved flexibility.For instance, a more advancedhtml() function could acceptadict of HTML attributes directly in the template:

attributes={'src':'limburger.jpg','alt':'lovely cheese'}template=t'<img{attributes}>'asserthtml(template)=='<img src="limburger.jpg" alt="lovely cheese" />'

Of course, template processing code does not need to return a string-like result.An evenmore advancedhtml() could return a custom type representinga DOM-like structure.

With t-strings in place, developers can write systems that sanitise SQL,make safe shell operations, improve logging, tackle modern ideas in webdevelopment (HTML, CSS, and so on), and implement lightweight custom business DSLs.

(Contributed by Jim Baker, Guido van Rossum, Paul Everitt, Koudai Aono,Lysandros Nikolaou, Dave Peck, Adam Turner, Jelle Zijlstra, Bénédikt Tran,and Pablo Galindo Salgado ingh-132661.)

See also

PEP 750.

PEP 768: Safe external debugger interface

Python 3.14 introduces a zero-overhead debugging interface that allowsdebuggers and profilers to safely attach to running Python processeswithout stopping or restarting them.This is a significant enhancement to Python’s debugging capabilities,meaning that unsafe alternatives are no longer required.

The new interface provides safe execution points for attaching debugger codewithout modifying the interpreter’s normal execution pathor adding any overhead at runtime.Due to this, tools can now inspect and interact with Python applicationsin real-time, which is a crucial capability for high-availability systemsand production environments.

For convenience, this interface is implemented in thesys.remote_exec()function. For example:

importsysfromtempfileimportNamedTemporaryFilewithNamedTemporaryFile(mode='w',suffix='.py',delete=False)asf:script_path=f.namef.write(f'import my_debugger; my_debugger.connect({os.getpid()})')# Execute in process with PID 1234print('Behold! An offering:')sys.remote_exec(1234,script_path)

This function allows sending Python code to be executed in a target processat the next safe execution point.However, tool authors can also implement the protocol directly as describedin the PEP, which details the underlying mechanisms used to safely attach torunning processes.

The debugging interface has been carefully designed with security in mindand includes several mechanisms to control access:

(Contributed by Pablo Galindo Salgado, Matt Wozniski, and Ivona Stojanovicingh-131591.)

See also

PEP 768.

A new type of interpreter

A new type of interpreter has been added to CPython.It uses tail calls between small C functions that implement individualPython opcodes, rather than one large Ccase statement.For certain newer compilers, this interpreter providessignificantly better performance. Preliminary benchmarks suggest a geometricmean of 3-5% faster on the standardpyperformance benchmark suite,depending on platform and architecture.The baseline is Python 3.14 built with Clang 19, without this new interpreter.

This interpreter currently only works with Clang 19 and neweron x86-64 and AArch64 architectures.However, a future release of GCC is expected to support this as well.

This feature is opt-in for now. Enabling profile-guided optimization is highlyrecommendeded when using the new interpreter as it is the only configurationthat has been tested and validated for improved performance.For further information, see--with-tail-call-interp.

Note

This is not to be confused withtail call optimization of Pythonfunctions, which is currently not implemented in CPython.

This new interpreter type is an internal implementation detail of the CPythoninterpreter. It doesn’t change the visible behavior of Python programs atall. It can improve their performance, but doesn’t change anything else.

(Contributed by Ken Jin ingh-128563, with ideas on how to implement thisin CPython by Mark Shannon, Garrett Gu, Haoran Xu, and Josh Haberman.)

Free-threaded mode improvements

CPython’s free-threaded mode (PEP 703), initially added in 3.13,has been significantly improved in Python 3.14.The implementation described in PEP 703 has been finished, including C APIchanges, and temporary workarounds in the interpreter were replaced withmore permanent solutions.The specializing adaptive interpreter (PEP 659) is now enabledin free-threaded mode, which along with many other optimizationsgreatly improves its performance.The performance penalty on single-threaded code in free-threaded modeis now roughly 5-10%, depending on the platform and C compiler used.

From Python 3.14, when compiling extension modules for the free-threaded build ofCPython on Windows, the preprocessor variablePy_GIL_DISABLED now needs tobe specified by the build backend, as it will no longer be determinedautomatically by the C compiler. For a running interpreter, the setting thatwas used at compile time can be found usingsysconfig.get_config_var().

The new-Xcontext_aware_warnings flag controls ifconcurrent safe warnings controlis enabled. The flag defaults to true for the free-threaded buildand false for the GIL-enabled build.

A newthread_inherit_context flag has been added,which if enabled means that threads created withthreading.Threadstart with a copy of theContext() of the caller ofstart(). Most significantly, this makes the warningfiltering context established bycatch_warnings be“inherited” by threads (or asyncio tasks) started within that context. It alsoaffects other modules that use context variables, such as thedecimalcontext manager.This flag defaults to true for the free-threaded build and false forthe GIL-enabled build.

(Contributed by Sam Gross, Matt Page, Neil Schemenauer, Thomas Wouters,Donghee Na, Kirill Podoprigora, Ken Jin, Itamar Oren, Brett Simmers,Dino Viehland, Nathan Goldbaum, Ralf Gommers, Lysandros Nikolaou, Kumar Aditya,Edgar Margffoy, and many others.Some of these contributors are employed by Meta, which has continued to providesignificant engineering resources to support this project.)

Improved error messages

  • The interpreter now provides helpful suggestions when it detects typos in Pythonkeywords. When a word that closely resembles a Python keyword is encountered,the interpreter will suggest the correct keyword in the error message. Thisfeature helps programmers quickly identify and fix common typing mistakes. Forexample:

    >>>whilleTrue:...passTraceback (most recent call last):  File"<stdin>", line1whilleTrue:^^^^^^SyntaxError:invalid syntax. Did you mean 'while'?

    While the feature focuses on the most common cases, some variations ofmisspellings may still result in regular syntax errors.(Contributed by Pablo Galindo ingh-132449.)

  • elif statements that follow anelse block now havea specific error message.(Contributed by Steele Farnsworth ingh-129902.)

    >>>ifwho=="me":...print("It's me!")...else:...print("It's not me!")...elifwhoisNone:...print("Who is it?")File "<stdin>", line 5  elif who is None:  ^^^^SyntaxError: 'elif' block follows an 'else' block
  • If a statement is passed to theConditional expressions afterelse,or one ofpass,break, orcontinueis passed beforeif, then theerror message highlights where theexpression isrequired. (Contributed by Sergey Miryanov ingh-129515.)

    >>>x=1ifTrueelsepassTraceback (most recent call last):  File"<string>", line1x=1ifTrueelsepass^^^^SyntaxError:expected expression after 'else', but statement is given>>>x=continueifTrueelsebreakTraceback (most recent call last):  File"<string>", line1x=continueifTrueelsebreak^^^^^^^^SyntaxError:expected expression before 'if', but statement is given
  • When incorrectly closed strings are detected, the error message suggeststhat the string may be intended to be part of the string.(Contributed by Pablo Galindo ingh-88535.)

    >>>"The interesting object "Theimportantobject" is very important"Traceback (most recent call last):SyntaxError:invalid syntax. Is this intended to be part of the string?
  • When strings have incompatible prefixes, the error now showswhich prefixes are incompatible.(Contributed by Nikita Sobolev ingh-133197.)

    >>>ub'abc'  File"<python-input-0>", line1ub'abc'^^SyntaxError:'u' and 'b' prefixes are incompatible
  • Improved error messages when usingas with incompatible targets in:

    • Imports:import...as...

    • From imports:from...import...as...

    • Except handlers:except...as...

    • Pattern-match cases:case...as...

    (Contributed by Nikita Sobolev ingh-123539,gh-123562, andgh-123440.)

  • Improved error message when trying to add an instance of an unhashable type toadict orset.(Contributed by CF Bolz-Tereick and Victor Stinner ingh-132828.)

    >>>s=set()>>>s.add({'pages':12,'grade':'A'})Traceback (most recent call last):  File"<python-input-1>", line1, in<module>s.add({'pages':12,'grade':'A'})~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^TypeError:cannot use 'dict' as a set element (unhashable type: 'dict')>>>d={}>>>l=[1,2,3]>>>d[l]=12Traceback (most recent call last):  File"<python-input-4>", line1, in<module>d[l]=12~^^^TypeError:cannot use 'list' as a dict key (unhashable type: 'list')
  • Improved error message when an object supporting the synchronouscontext manager protocol is entered usingasyncwithinstead ofwith,and vice versa for the asynchronous context manager protocol.(Contributed by Bénédikt Tran ingh-128398.)

PEP 784: Zstandard support in the standard library

The newcompression package contains modulescompression.lzma,compression.bz2,compression.gzip andcompression.zlibwhich re-export thelzma,bz2,gzip andzlibmodules respectively. The new import names undercompression are thepreferred names for importing these compression modules from Python 3.14. However,the existing modules names have not been deprecated. Any deprecation or removalof the existing compression modules will occur no sooner than five years afterthe release of 3.14.

The newcompression.zstd module provides compression and decompressionAPIs for the Zstandard format via bindings toMeta’s zstd library. Zstandard is a widely adopted, highlyefficient, and fast compression format. In addition to the APIs introduced incompression.zstd, support for reading and writing Zstandard compressedarchives has been added to thetarfile,zipfile, andshutil modules.

Here’s an example of using the new module to compress some data:

fromcompressionimportzstdimportmathdata=str(math.pi).encode()*20compressed=zstd.compress(data)ratio=len(compressed)/len(data)print(f"Achieved compression ratio of{ratio}")

As can be seen, the API is similar to the APIs of thelzma andbz2 modules.

(Contributed by Emma Harper Smith, Adam Turner, Gregory P. Smith, Tomas Roun,Victor Stinner, and Rogdham ingh-132983.)

See also

PEP 784.

Asyncio introspection capabilities

Added a new command-line interface to inspect running Python processesusing asynchronous tasks, available viapython-masynciopsPIDorpython-masynciopstreePID.

Theps subcommand inspects the given process ID (PID) and displaysinformation about currently running asyncio tasks.It outputs a task table: a flat listing of all tasks, their names,their coroutine stacks, and which tasks are awaiting them.

Thepstree subcommand fetches the same information, but instead renders avisual async call tree, showing coroutine relationships in a hierarchical format.This command is particularly useful for debugging long-running or stuckasynchronous programs.It can help developers quickly identify where a program is blocked,what tasks are pending, and how coroutines are chained together.

For example given this code:

importasyncioasyncdefplay_track(track):awaitasyncio.sleep(5)print(f'🎵 Finished:{track}')asyncdefplay_album(name,tracks):asyncwithasyncio.TaskGroup()astg:fortrackintracks:tg.create_task(play_track(track),name=track)asyncdefmain():asyncwithasyncio.TaskGroup()astg:tg.create_task(play_album('Sundowning',['TNDNBTG','Levitate']),name='Sundowning')tg.create_task(play_album('TMBTE',['DYWTYLM','Aqua Regia']),name='TMBTE')if__name__=='__main__':asyncio.run(main())

Executing the new tool on the running process will yield a table like this:

python-masynciops12345tidtaskidtasknamecoroutinestackawaiterchainawaiternameawaiterid------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------19355000x7fc930c18050Task-1TaskGroup._aexit->TaskGroup.__aexit__->main0x019355000x7fc930c18230SundowningTaskGroup._aexit->TaskGroup.__aexit__->albumTaskGroup._aexit->TaskGroup.__aexit__->mainTask-10x7fc930c1805019355000x7fc93173fa50TMBTETaskGroup._aexit->TaskGroup.__aexit__->albumTaskGroup._aexit->TaskGroup.__aexit__->mainTask-10x7fc930c1805019355000x7fc93173fdf0TNDNBTGsleep->playTaskGroup._aexit->TaskGroup.__aexit__->albumSundowning0x7fc930c1823019355000x7fc930d32510Levitatesleep->playTaskGroup._aexit->TaskGroup.__aexit__->albumSundowning0x7fc930c1823019355000x7fc930d32890DYWTYLMsleep->playTaskGroup._aexit->TaskGroup.__aexit__->albumTMBTE0x7fc93173fa5019355000x7fc93161ec30AquaRegiasleep->playTaskGroup._aexit->TaskGroup.__aexit__->albumTMBTE0x7fc93173fa50

or a tree like this:

python-masynciopstree12345└──(T)Task-1└──mainexample.py:13└──TaskGroup.__aexit__Lib/asyncio/taskgroups.py:72└──TaskGroup._aexitLib/asyncio/taskgroups.py:121├──(T)Sundowning└──albumexample.py:8└──TaskGroup.__aexit__Lib/asyncio/taskgroups.py:72└──TaskGroup._aexitLib/asyncio/taskgroups.py:121├──(T)TNDNBTG└──playexample.py:4└──sleepLib/asyncio/tasks.py:702└──(T)Levitate└──playexample.py:4└──sleepLib/asyncio/tasks.py:702└──(T)TMBTE└──albumexample.py:8└──TaskGroup.__aexit__Lib/asyncio/taskgroups.py:72└──TaskGroup._aexitLib/asyncio/taskgroups.py:121├──(T)DYWTYLM└──playexample.py:4└──sleepLib/asyncio/tasks.py:702└──(T)AquaRegia└──playexample.py:4└──sleepLib/asyncio/tasks.py:702

If a cycle is detected in the async await graph (which could indicate aprogramming issue), the tool raises an error and lists the cycle paths thatprevent tree construction:

python-masynciopstree12345ERROR:await-graphcontainscycles-cannotprintatree!cycle:Task-2Task-3Task-2

(Contributed by Pablo Galindo, Łukasz Langa, Yury Selivanov, and MartaGomez Macias ingh-91048.)

Concurrent safe warnings control

Thewarnings.catch_warnings context manager will now optionallyuse a context variable for warning filters. This is enabled by settingthecontext_aware_warnings flag, either with the-Xcommand-line option or an environment variable. This gives predictablewarnings control when usingcatch_warnings combined withmultiple threads or asynchronous tasks. The flag defaults to true for thefree-threaded build and false for the GIL-enabled build.

(Contributed by Neil Schemenauer and Kumar Aditya ingh-130010.)

Other language changes

  • All Windows code pages are now supported as ‘cpXXX’ codecs on Windows.(Contributed by Serhiy Storchaka ingh-123803.)

  • Implement mixed-mode arithmetic rules combining real and complex numbersas specified by the C standard since C99.(Contributed by Sergey B Kirpichev ingh-69639.)

  • More syntax errors are now detected regardless of optimisation andthe-O command-line option.This includes writes to__debug__, incorrect use ofawait,and asynchronous comprehensions outside asynchronous functions.For example,python-O-c'assert(__debug__:=1)'orpython-O-c'assertawait1' now produceSyntaxErrors.(Contributed by Irit Katriel and Jelle Zijlstra ingh-122245 &gh-121637.)

  • When subclassing a pure C type, the C slots for the new typeare no longer replaced with a wrapped version on class creationif they are not explicitly overridden in the subclass.(Contributed by Tomasz Pytel ingh-132284.)

Built-ins

Command line and environment

  • The import time flag can now track modules that are already loaded (‘cached’),via the new-Ximporttime=2.When such a module is imported, theself andcumulative timesare replaced by the stringcached.

    Values above2 for-Ximporttime are now reserved for future use.

    (Contributed by Noah Kim and Adam Turner ingh-118655.)

  • The command-line option-c now automatically dedents its codeargument before execution. The auto-dedentation behavior mirrorstextwrap.dedent().(Contributed by Jon Crall and Steven Sun ingh-103998.)

  • -J is no longer a reserved flag forJython,and now has no special meaning.(Contributed by Adam Turner ingh-133336.)

PEP 758: Allowexcept andexcept* expressions without brackets

Theexcept andexcept* expressionsnow allow brackets to be omitted when there are multiple exception typesand theas clause is not used.For example:

try:connect_to_server()exceptTimeoutError,ConnectionRefusedError:print('The network has ceased to be!')

(Contributed by Pablo Galindo and Brett Cannon inPEP 758 andgh-131831.)

PEP 765: Control flow infinally blocks

The compiler now emits aSyntaxWarning when areturn,break, orcontinue statement have the effect ofleaving afinally block.This change is specified inPEP 765.

In situations where this change is inconvenient (such as those where thewarnings are redundant due to code linting), thewarning filter can be used to turn off all syntax warnings by addingignore::SyntaxWarning as a filter. This can be specified in combinationwith a filter that converts other warnings to errors (for example, passing-Werror-Wignore::SyntaxWarning as CLI options, or settingPYTHONWARNINGS=error,ignore::SyntaxWarning).

Note that applying such a filter at runtime using thewarnings modulewill only suppress the warning in code that is compiledafter the filter isadjusted. Code that is compiled prior to the filter adjustment (for example,when a module is imported) will still emit the syntax warning.

(Contributed by Irit Katriel ingh-130080.)

Incremental garbage collection

The cycle garbage collector is now incremental.This means that maximum pause times are reducedby an order of magnitude or more for larger heaps.

There are now only two generations: young and old.Whengc.collect() is not called directly, theGC is invoked a little less frequently. When invoked, itcollects the young generation and an increment of theold generation, instead of collecting one or more generations.

The behavior ofgc.collect() changes slightly:

  • gc.collect(1): Performs an increment of garbage collection,rather than collecting generation 1.

  • Other calls togc.collect() are unchanged.

(Contributed by Mark Shannon ingh-108362.)

Default interactive shell

  • The defaultinteractive shell now highlights Python syntax.The feature is enabled by default, save ifPYTHON_BASIC_REPLor any other environment variable that disables colour is set.SeeControlling color for details.

    The default color theme for syntax highlighting strives for good contrastand exclusively uses the 4-bit VGA standard ANSI color codes for maximumcompatibility. The theme can be customized using an experimental API_colorize.set_theme().This can be called interactively or in thePYTHONSTARTUP script.Note that this function has no stability guarantees,and may change or be removed.

    (Contributed by Łukasz Langa ingh-131507.)

  • The defaultinteractive shell now supports import auto-completion.This means that typingimportco and pressing<Tab> will suggestmodules starting withco. Similarly, typingfromconcurrentimportiwill suggest submodules ofconcurrent starting withi.Note that autocompletion of module attributes is not currently supported.(Contributed by Tomas Roun ingh-69605.)

New modules

  • annotationlib:For introspectingannotations.SeePEP 749 for more details.(Contributed by Jelle Zijlstra ingh-119180.)

  • compression (includingcompression.zstd):A package for compression-related modules,including a new module to support the Zstandard compression format.SeePEP 784 for more details.(Contributed by Emma Harper Smith, Adam Turner, Gregory P. Smith, Tomas Roun,Victor Stinner, and Rogdham ingh-132983.)

  • concurrent.interpreters:Support for multiple interpreters in the standard library.SeePEP 734 for more details.(Contributed by Eric Snow ingh-134939.)

  • string.templatelib:Support for template string literals (t-strings).SeePEP 750 for more details.(Contributed by Jim Baker, Guido van Rossum, Paul Everitt, Koudai Aono,Lysandros Nikolaou, Dave Peck, Adam Turner, Jelle Zijlstra, Bénédikt Tran,and Pablo Galindo Salgado ingh-132661.)

Improved modules

argparse

ast

  • Addcompare(), a function for comparing two ASTs.(Contributed by Batuhan Taskaya and Jeremy Hylton ingh-60191.)

  • Add support forcopy.replace() for AST nodes.(Contributed by Bénédikt Tran ingh-121141.)

  • Docstrings are now removed from an optimized AST in optimization level 2.(Contributed by Irit Katriel ingh-123958.)

  • Therepr() output for AST nodes now includes more information.(Contributed by Tomas Roun ingh-116022.)

  • When called with an AST as input, theparse() functionnow always verifies that the root node type is appropriate.(Contributed by Irit Katriel ingh-130139.)

  • Add new options to the command-line interface:--feature-version,--optimize, and--show-empty.(Contributed by Semyon Moroz ingh-133367.)

asyncio

calendar

concurrent.futures

  • Add a new executor class,InterpreterPoolExecutor,which exposes multiple Python interpreters in the same process(‘subinterpreters’) to Python code.This uses a pool of independent Python interpreters to execute callsasynchronously.

    This is separate from the newinterpreters moduleintroduced byPEP 734.(Contributed by Eric Snow ingh-124548.)

  • On Unix platforms other than macOS,‘forkserver’ is now the the defaultstartmethod forProcessPoolExecutor(replacing‘fork’).This change does not affect Windows or macOS, where‘spawn’ remains the default start method.

    If the threading incompatiblefork method is required, you must explicitlyrequest it by supplying a multiprocessing contextmp_context toProcessPoolExecutor.

    Seeforkserver restrictionsfor information and differences with thefork method and how this changemay affect existing code with mutable global shared variables and/or sharedobjects that can not be automaticallypickled.

    (Contributed by Gregory P. Smith ingh-84559.)

  • Add two new methods toProcessPoolExecutor,terminate_workers()andkill_workers(),as ways to terminate or kill all living worker processes in the given pool.(Contributed by Charles Machalow ingh-130849.)

  • Add the optionalbuffersize parameter toExecutor.map to limit the number of submittedtasks whose results have not yet been yielded. If the buffer is full,iteration over theiterables pauses until a result is yielded from thebuffer.(Contributed by Enzo Bonnal and Josh Rosenberg ingh-74028.)

configparser

  • configparser will no longer write config files it cannot read,to improve security.Attempting towrite() keys containingdelimiters or beginning with the section header pattern will raise anInvalidWriteError.(Contributed by Jacob Lincoln ingh-129270.)

contextvars

ctypes

curses

datetime

decimal

difflib

  • Comparison pages with highlighted changes generated by theHtmlDiff class now support ‘dark mode’.(Contributed by Jiahao Li ingh-129939.)

dis

errno

faulthandler

fnmatch

  • Addfilterfalse(), a function to reject namesmatching a given pattern.(Contributed by Bénédikt Tran ingh-74598.)

fractions

functools

getopt

  • Add support for options with optional arguments.(Contributed by Serhiy Storchaka ingh-126374.)

  • Add support for returning intermixed options and non-option arguments in order.(Contributed by Serhiy Storchaka ingh-126390.)

getpass

  • Support keyboard feedback in thegetpass() function viathe keyword-only optional argumentecho_char.Placeholder characters are rendered whenever a character is entered,and removed when a character is deleted.(Contributed by Semyon Moroz ingh-77065.)

graphlib

heapq

hmac

  • Add a built-in implementation for HMAC (RFC 2104) using formally verifiedcode from theHACL* project.This implementation is used as a fallback when the OpenSSL implementationof HMAC is not available.(Contributed by Bénédikt Tran ingh-99108.)

http

imaplib

inspect

io

json

linecache

  • getline() can now retrieve source code for frozen modules.(Contributed by Tian Gao ingh-131638.)

logging.handlers

math

  • Added more detailed error messages for domain errors in the module.(Contributed by Charlie Zhao and Sergey B Kirpichev ingh-101410.)

mimetypes

  • Add a publiccommand-line for the module,invoked viapython -m mimetypes.(Contributed by Oleg Iarygin and Hugo van Kemenade ingh-93096.)

  • Add several new MIME types based on RFCs and common usage:

    Microsoft andRFC 8081 MIME types for fonts

    • Embedded OpenType:application/vnd.ms-fontobject

    • OpenType Layout (OTF)font/otf

    • TrueType:font/ttf

    • WOFF 1.0font/woff

    • WOFF 2.0font/woff2

    RFC 9559 MIME types for Matroska audiovisualdata container structures

    • audio with no video:audio/matroska (.mka)

    • video:video/matroska (.mkv)

    • stereoscopic video:video/matroska-3d (.mk3d)

    Images with RFCs

    • RFC 1494: CCITT Group 3 (.g3)

    • RFC 3362: Real-time Facsimile, T.38 (.t38)

    • RFC 3745: JPEG 2000 (.jp2), extension (.jpx) and compound (.jpm)

    • RFC 3950: Tag Image File Format Fax eXtended, TIFF-FX (.tfx)

    • RFC 4047: Flexible Image Transport System (.fits)

    • RFC 7903: Enhanced Metafile (.emf) and Windows Metafile (.wmf)

    Other MIME type additions and changes

    • RFC 2361: Change type for.avi tovideo/vnd.aviand for.wav toaudio/vnd.wave

    • RFC 4337: Add MPEG-4audio/mp4 (.m4a)

    • RFC 5334: Add Ogg media (.oga,.ogg and.ogx)

    • RFC 6713: Add gzipapplication/gzip (.gz)

    • RFC 9639: Add FLACaudio/flac (.flac)

    • RFC 9512application/yaml MIME type for YAML files (.yamland.yml)

    • Add 7zapplication/x-7z-compressed (.7z)

    • Add Android Packageapplication/vnd.android.package-archive (.apk)when not strict

    • Add debapplication/x-debian-package (.deb)

    • Add glTF binarymodel/gltf-binary (.glb)

    • Add glTF JSON/ASCIImodel/gltf+json (.gltf)

    • Add M4Vvideo/x-m4v (.m4v)

    • Add PHPapplication/x-httpd-php (.php)

    • Add RARapplication/vnd.rar (.rar)

    • Add RPMapplication/x-rpm (.rpm)

    • Add STLmodel/stl (.stl)

    • Add Windows Media Videovideo/x-ms-wmv (.wmv)

    • De facto: Add WebMaudio/webm (.weba)

    • ECMA-376:Add.docx,.pptx and.xlsx types

    • OASIS:Add OpenDocument.odg,.odp,.ods and.odt types

    • W3C:Add EPUBapplication/epub+zip (.epub)

    (Contributed by Sahil Prajapati and Hugo van Kemenade ingh-84852,by Sasha “Nelie” Chernykh and Hugo van Kemenade ingh-132056,and by Hugo van Kemenade ingh-89416,gh-85957, andgh-129965.)

multiprocessing

  • On Unix platforms other than macOS,‘forkserver’ is now the the defaultstartmethod(replacing‘fork’).This change does not affect Windows or macOS, where‘spawn’ remains the default start method.

    If the threading incompatiblefork method is required, you must explicitlyrequest it via a context fromget_context() (preferred)or change the default viaset_start_method().

    Seeforkserver restrictionsfor information and differences with thefork method and how this changemay affect existing code with mutable global shared variables and/or sharedobjects that can not be automaticallypickled.

    (Contributed by Gregory P. Smith ingh-84559.)

  • multiprocessing’s'forkserver' start method now authenticatesits control socket to avoid solely relying on filesystem permissionsto restrict what other processes could cause the forkserver to spawn workersand run code.(Contributed by Gregory P. Smith forgh-97514.)

  • Themultiprocessing proxy objectsforlist anddict types gain previously overlooked missing methods:

    • clear() andcopy() for proxies oflist

    • fromkeys(),reversed(d),d|{},{}|d,d|={'b':2} for proxies ofdict

    (Contributed by Roy Hyunjin Han forgh-103134.)

  • Add support for sharedset objects viaSyncManager.set().Theset() inManager() method is now available.(Contributed by Mingyu Park ingh-129949.)

  • Add theinterrupt()tomultiprocessing.Process objects, which terminates the childprocess by sendingSIGINT. This enablesfinally clauses to print a stack trace for the terminatedprocess. (Contributed by Artem Pulkin ingh-131913.)

operator

  • Addis_none() andis_not_none() as a pairof functions, such thatoperator.is_none(obj) is equivalenttoobjisNone andoperator.is_not_none(obj) is equivalenttoobjisnotNone.(Contributed by Raymond Hettinger and Nico Mexis ingh-115808.)

os

os.path

pathlib

  • Add methods topathlib.Path to recursively copy or move files anddirectories:

    • copy() copies a file or directory tree to a destination.

    • copy_into() copiesinto a destination directory.

    • move() moves a file or directory tree to a destination.

    • move_into() movesinto a destination directory.

    (Contributed by Barney Gale ingh-73991.)

  • Add theinfo attribute, which stores an objectimplementing the newpathlib.types.PathInfo protocol. Theobject supports querying the file type and internally cachingstat() results. Path objects generated byiterdir() are initialized with file type informationgleaned from scanning the parent directory.(Contributed by Barney Gale ingh-125413.)

pdb

  • Thepdb module now supports remote attaching to a running Python processusing a new-pPID command-line option:

    python-mpdb-p1234

    This will connect to the Python process with the given PID and allow you todebug it interactively. Notice that due to how the Python interpreter worksattaching to a remote process that is blocked in a system call or waiting forI/O will only work once the next bytecode instruction is executed or when theprocess receives a signal.

    This feature usesPEP 768and the newsys.remote_exec() function to attach to the remote processand send the PDB commands to it.

    (Contributed by Matt Wozniski and Pablo Galindo ingh-131591.)

  • Hardcoded breakpoints (breakpoint() andset_trace()) nowreuse the most recentPdb instance that callsset_trace(), instead of creating a new one each time.As a result, all the instance specific data likedisplay andcommands are preserved across hardcoded breakpoints.(Contributed by Tian Gao ingh-121450.)

  • Add a new argumentmode topdb.Pdb. Disable therestartcommand whenpdb is ininline mode.(Contributed by Tian Gao ingh-123757.)

  • A confirmation prompt will be shown when the user tries to quitpdbininline mode.y,Y,<Enter> orEOF will confirmthe quit and callsys.exit(), instead of raisingbdb.BdbQuit.(Contributed by Tian Gao ingh-124704.)

  • Inline breakpoints likebreakpoint() orpdb.set_trace() willalways stop the program at calling frame, ignoring theskip pattern(if any).(Contributed by Tian Gao ingh-130493.)

  • <tab> at the beginning of the line inpdb multi-line input willfill in a 4-space indentation now, instead of inserting a\t character.(Contributed by Tian Gao ingh-130471.)

  • Auto-indent is introduced inpdb multi-line input. It will eitherkeep the indentation of the last line or insert a 4-space indentation whenit detects a new code block.(Contributed by Tian Gao ingh-133350.)

  • $_asynctask is added to access the current asyncio task if applicable.(Contributed by Tian Gao ingh-124367.)

  • pdb.set_trace_async() is added to support debugging asynciocoroutines.await statements are supported with thisfunction.(Contributed by Tian Gao ingh-132576.)

  • Source code displayed inpdb will be syntax-highlighted. This featurecan be controlled using the same methods as the defaultinteractiveshell, in addition to the newly addedcolorize argument ofpdb.Pdb.(Contributed by Tian Gao and Łukasz Langa ingh-133355.)

pickle

  • Set the default protocol version on thepickle module to 5.For more details, seepickle protocols.

  • Add exception notes for pickle serialization errors that allowidentifying the source of the error.(Contributed by Serhiy Storchaka ingh-122213.)

platform

pydoc

  • Annotations in help output are now usuallydisplayed in a format closer to that in the original source.(Contributed by Jelle Zijlstra ingh-101552.)

re

  • Support\z as a synonym for\Z inregularexpressions.It is interpreted unambiguously in many other regular expression engines,unlike\Z, which has subtly different behavior.(Contributed by Serhiy Storchaka ingh-133306.)

  • \B inregularexpression now matches the empty input string,meaning that it is now always the opposite of\b.(Contributed by Serhiy Storchaka ingh-124130.)

socket

  • Improve and fix support for Bluetooth sockets.

ssl

  • Indicate through theHAS_PHA Boolean whether thesslmodule supports TLSv1.3 post-handshake client authentication (PHA).(Contributed by Will Childs-Klein ingh-128036.)

struct

  • Support thefloatcomplex anddoublecomplex C types inthestruct module (formatting characters'F' and'D'respectively).(Contributed by Sergey B Kirpichev ingh-121249.)

symtable

sys

  • The previously undocumented special functionsys.getobjects(),which only exists in specialized builds of Python, may now return objectsfrom other interpreters than the one it’s called in.(Contributed by Eric Snow ingh-125286.)

  • Addsys._is_immortal() for determining if an object isimmortal.(Contributed by Peter Bierma ingh-128509.)

  • On FreeBSD,sys.platform no longer contains the major version number.It is always'freebsd', instead of'freebsd13' or'freebsd14'.(Contributed by Michael Osipov ingh-129393.)

  • RaiseDeprecationWarning forsys._clear_type_cache(). Thisfunction was deprecated in Python 3.13 but it didn’t raise a runtime warning.

  • Addsys.remote_exec() to implement the new external debugger interface.SeePEP 768 for details.(Contributed by Pablo Galindo Salgado, Matt Wozniski, and Ivona Stojanovicingh-131591.)

  • Add thesys._jit namespace, containing utilities for introspectingjust-in-time compilation.(Contributed by Brandt Bucher ingh-133231.)

sys.monitoring

sysconfig

tarfile

threading

tkinter

  • Maketkinter widget methodsafter() andafter_idle()accept keyword arguments.(Contributed by Zhikang Yan ingh-126899.)

  • Add ability to specify a name fortkinter.OptionMenu andtkinter.ttk.OptionMenu.(Contributed by Zhikang Yan ingh-130482.)

turtle

types

typing

  • Thetypes.UnionType andtyping.Union types are nowaliases for each other, meaning that both old-style unions(created withUnion[int,str]) and new-style unions (int|str)now create instances of the same runtime type. This unifies the behaviorbetween the two syntaxes, but leads to some differences in behavior thatmay affect users who introspect types at runtime:

    • Both syntaxes for creating a union now produce the same stringrepresentation inrepr().For example,repr(Union[int,str]) is now"int|str" instead of"typing.Union[int,str]".

    • Unions created using the old syntax are no longer cached.Previously, runningUnion[int,str] multiple times would returnthe same object (Union[int,str]isUnion[int,str] would beTrue),but now it will return two different objects.Use== to compare unions for equality, notis.New-style unions have never been cached this way.This change could increase memory usage for some programs that usea large number of unions created by subscriptingtyping.Union.However, several factors offset this cost:unions used in annotations are no longer evaluated by default in Python3.14 because ofPEP 649; an instance oftypes.UnionType isitself much smaller than the object returned byUnion[] was on priorPython versions; and removing the cache also saves some space.It is therefore unlikely that this change will cause a significant increasein memory usage for most users.

    • Previously, old-style unions were implemented using the private classtyping._UnionGenericAlias.This class is no longer needed for the implementation,but it has been retained for backward compatibility,with removal scheduled for Python 3.17.Users should use documented introspection helpers likeget_origin() andtyping.get_args() instead ofrelying on private implementation details.

    • It is now possible to usetyping.Union itself inisinstance() checks.For example,isinstance(int|str,typing.Union) will returnTrue;previously this raisedTypeError.

    • The__args__ attribute oftyping.Union objects isno longer writable.

    • It is no longer possible to set any attributes onUnionobjects.This only ever worked for dunder attributes on previous versions, was neverdocumented to work, and was subtly broken in many cases.

    (Contributed by Jelle Zijlstra ingh-105499.)

  • TypeAliasType now supports star unpacking.

unicodedata

  • The Unicode database has been updated to Unicode 16.0.0.

unittest

urllib

  • Upgrade HTTP digest authentication algorithm forurllib.request bysupporting SHA-256 digest authentication as specified inRFC 7616.(Contributed by Calvin Bui ingh-128193.)

  • Improve ergonomics and standards compliance when parsing and emittingfile: URLs.

    Inurl2pathname():

    • Accept a complete URL when the newrequire_scheme argument is set totrue.

    • Discard URL authority if it matches the local hostname.

    • Discard URL authority if it resolves to a local IP address when the newresolve_host argument is set to true.

    • Discard URL query and fragment components.

    • RaiseURLError if a URL authority isn’t local,except on Windows where we return a UNC path as before.

    Inpathname2url():

    • Return a complete URL when the newadd_scheme argument is set to true.

    • Include an empty URL authority when a path begins with a slash. Forexample, the path/etc/hosts is converted to the URL///etc/hosts.

    On Windows, drive letters are no longer converted to uppercase, and:characters not following a drive letter no longer cause anOSErrorexception to be raised.

    (Contributed by Barney Gale ingh-125866.)

uuid

webbrowser

  • Names in theBROWSER environment variable can now refer to alreadyregistered browsers for thewebbrowser module, instead of alwaysgenerating a new browser command.

    This makes it possible to setBROWSER to the value of one of thesupported browsers on macOS.

zipfile

Optimizations

asyncio

  • Standard benchmark results have improved by 10-20% following theimplementation of a new per-thread doubly linked listfornativetasks,also reducing memory usage.This enables external introspection tools such aspython -m asyncio pstreeto introspect the call graph of asyncio tasks running in all threads.(Contributed by Kumar Aditya ingh-107803.)

  • The module now has first class support forfree-threading builds.This enables parallel execution of multiple event loops acrossdifferent threads, scaling linearly with the number of threads.(Contributed by Kumar Aditya ingh-128002.)

base64

  • b16decode() is now up to six times faster.(Contributed by Bénédikt Tran, Chris Markiewicz, and Adam Turneringh-118761.)

bdb

  • The basic debugger now has asys.monitoring-based backend,which can be selected via the passing'monitoring'to theBdb class’s newbackend parameter.(Contributed by Tian Gao ingh-124533.)

difflib

gc

  • The newincremental garbage collectormeans that maximum pause times are reducedby an order of magnitude or more for larger heaps.

    Because of this optimization, the meaning of the results ofget_threshold() andset_threshold() have changed,along withget_count() andget_stats().

    • For backwards compatibility,get_threshold() continues to returna three-item tuple.The first value is the threshold for young collections, as before;the second value determines the rate at which the old collection is scanned(the default is 10, and higher values mean that the old collectionis scanned more slowly).The third value is now meaningless and is always zero.

    • set_threshold() now ignores any items after the second.

    • get_count() andget_stats() continue to returnthe same format of results.The only difference is that instead of the results referring tothe young, aging and old generations,the results refer to the young generationand the aging and collecting spaces of the old generation.

    In summary, code that attempted to manipulate the behavior of the cycle GCmay not work exactly as intended, but it is very unlikely to be harmful.All other code will work just fine.

    (Contributed by Mark Shannon ingh-108362.)

io

  • Opening and reading files now executes fewer system calls.Reading a small operating system cached file in full is up to 15% faster.(Contributed by Cody Maloney and Victor Stinneringh-120754 andgh-90102.)

pathlib

  • Path.read_bytes now uses unbuffered modeto open files, which is between 9% and 17% faster to read in full.(Contributed by Cody Maloney ingh-120754.)

pdb

uuid

  • uuid3() anduuid5() are now both roughly 40% fasterfor 16-byte names and 20% faster for 1024-byte names.Performance for longer names remains unchanged.(Contributed by Bénédikt Tran ingh-128150.)

  • uuid4() is now c. 30% faster.(Contributed by Bénédikt Tran ingh-128150.)

zlib

  • On Windows,zlib-ngis now used as the implementation of thezlib modulein the default binaries.There are no known incompatibilities betweenzlib-ngand the previously-usedzlib implementation.This should result in better performance at all compression levels.

    It is worth noting thatzlib.Z_BEST_SPEED (1) may result insignificantly less compression than the previous implementation,whilst also significantly reducing the time taken to compress.

    (Contributed by Steve Dower ingh-91349.)

Removed

argparse

  • Remove thetype,choices, andmetavar parametersofBooleanOptionalAction.These have been deprecated since Python 3.12.(Contributed by Nikita Sobolev ingh-118805.)

  • Callingadd_argument_group()on an argument group now raises aValueError.Similarly,add_argument_group()oradd_mutually_exclusive_group()on a mutually exclusive group now both raiseValueErrors.This ‘nesting’ was never supported, often failed to work correctly,and was unintentionally exposed through inheritance.This functionality has been deprecated since Python 3.11.(Contributed by Savannah Ostrowski ingh-127186.)

ast

  • Remove the following classes, which have been deprecated aliases ofConstant since Python 3.8 and have emitteddeprecation warnings since Python 3.12:

    • Bytes

    • Ellipsis

    • NameConstant

    • Num

    • Str

    As a consequence of these removals, user-definedvisit_Num,visit_Str,visit_Bytes,visit_NameConstant andvisit_Ellipsis methodson customNodeVisitor subclasses will no longer be calledwhen theNodeVisitor subclass is visiting an AST.Define avisit_Constant method instead.

    (Contributed by Alex Waygood ingh-119562.)

  • Remove the following deprecated properties onast.Constant,which were present for compatibility with the now-removed AST classes:

    • Constant.n

    • Constant.s

    UseConstant.value instead.(Contributed by Alex Waygood ingh-119562.)

asyncio

  • Remove the following classes, methods, and functions,which have been deprecated since Python 3.12:

    • AbstractChildWatcher

    • FastChildWatcher

    • MultiLoopChildWatcher

    • PidfdChildWatcher

    • SafeChildWatcher

    • ThreadedChildWatcher

    • AbstractEventLoopPolicy.get_child_watcher()

    • AbstractEventLoopPolicy.set_child_watcher()

    • get_child_watcher()

    • set_child_watcher()

    (Contributed by Kumar Aditya ingh-120804.)

  • asyncio.get_event_loop() now raises aRuntimeErrorif there is no current event loop,and no longer implicitly creates an event loop.

    (Contributed by Kumar Aditya ingh-126353.)

    There’s a few patterns that useasyncio.get_event_loop(), mostof them can be replaced withasyncio.run().

    If you’re running an async function, simply useasyncio.run().

    Before:

    asyncdefmain():...loop=asyncio.get_event_loop()try:loop.run_until_complete(main())finally:loop.close()

    After:

    asyncdefmain():...asyncio.run(main())

    If you need to start something, for example, a server listening on a socketand then run forever, useasyncio.run() and anasyncio.Event.

    Before:

    defstart_server(loop):...loop=asyncio.get_event_loop()try:start_server(loop)loop.run_forever()finally:loop.close()

    After:

    defstart_server(loop):...asyncdefmain():start_server(asyncio.get_running_loop())awaitasyncio.Event().wait()asyncio.run(main())

    If you need to run something in an event loop, then run some blockingcode around it, useasyncio.Runner.

    Before:

    asyncdefoperation_one():...defblocking_code():...asyncdefoperation_two():...loop=asyncio.get_event_loop()try:loop.run_until_complete(operation_one())blocking_code()loop.run_until_complete(operation_two())finally:loop.close()

    After:

    asyncdefoperation_one():...defblocking_code():...asyncdefoperation_two():...withasyncio.Runner()asrunner:runner.run(operation_one())blocking_code()runner.run(operation_two())

email

importlib.abc

itertools

  • Remove support for copy, deepcopy, and pickle operationsfromitertools iterators.These have emitted aDeprecationWarning since Python 3.12.(Contributed by Raymond Hettinger ingh-101588.)

pathlib

  • Remove support for passing additional keyword argumentstoPath.In previous versions, any such arguments are ignored.(Contributed by Barney Gale ingh-74033.)

  • Remove support for passing additional positional arguments toPurePath.relative_to() andis_relative_to().In previous versions, any such arguments are joined ontoother.(Contributed by Barney Gale ingh-78707.)

pkgutil

  • Remove theget_loader() andfind_loader() functions,which have been deprecated since Python 3.12.(Contributed by Bénédikt Tran ingh-97850.)

pty

  • Remove themaster_open() andslave_open() functions,which have been deprecated since Python 3.12.Usepty.openpty() instead.(Contributed by Nikita Sobolev ingh-118824.)

sqlite3

urllib

  • Remove theQuoter class fromurllib.parse,which has been deprecated since Python 3.11.(Contributed by Nikita Sobolev ingh-118827.)

  • Remove theURLopener andFancyURLopener classesfromurllib.request,which have been deprecated since Python 3.3.

    myopener.open() can be replaced withurlopen().myopener.retrieve() can be replaced withurlretrieve().Customisations to the opener classes can be replaced by passingcustomized handlers tobuild_opener().(Contributed by Barney Gale ingh-84850.)

Deprecated

New deprecations

Pending removal in Python 3.15

  • The import system:

    • Setting__cached__ on a module whilefailing to set__spec__.cachedis deprecated. In Python 3.15,__cached__ will cease to be set ortake into consideration by the import system or standard library. (gh-97879)

    • Setting__package__ on a module whilefailing to set__spec__.parentis deprecated. In Python 3.15,__package__ will cease to be set ortake into consideration by the import system or standard library. (gh-97879)

  • ctypes:

    • The undocumentedctypes.SetPointerType() functionhas been deprecated since Python 3.13.

  • http.server:

    • The obsolete and rarely usedCGIHTTPRequestHandlerhas been deprecated since Python 3.13.No direct replacement exists.Anything is better than CGI to interfacea web server with a request handler.

    • The--cgi flag to thepython -m http.servercommand-line interface has been deprecated since Python 3.13.

  • importlib:

    • load_module() method: useexec_module() instead.

  • locale:

  • pathlib:

    • .PurePath.is_reserved()has been deprecated since Python 3.13.Useos.path.isreserved() to detect reserved paths on Windows.

  • platform:

    • platform.java_ver() has been deprecated since Python 3.13.This function is only useful for Jython support, has a confusing API,and is largely untested.

  • sysconfig:

  • threading:

    • RLock() will take no arguments in Python 3.15.Passing any arguments has been deprecated since Python 3.14,as the Python version does not permit any arguments,but the C version allows any number of positional or keyword arguments,ignoring every argument.

  • types:

  • typing:

    • The undocumented keyword argument syntax for creatingNamedTuple classes(for example,Point=NamedTuple("Point",x=int,y=int))has been deprecated since Python 3.13.Use the class-based syntax or the functional syntax instead.

    • When using the functional syntax ofTypedDicts, failingto pass a value to thefields parameter (TD=TypedDict("TD")) orpassingNone (TD=TypedDict("TD",None)) has been deprecatedsince Python 3.13.UseclassTD(TypedDict):pass orTD=TypedDict("TD",{})to create a TypedDict with zero field.

    • Thetyping.no_type_check_decorator() decorator functionhas been deprecated since Python 3.13.After eight years in thetyping module,it has yet to be supported by any major type checker.

  • sre_compile,sre_constants andsre_parse modules.

  • wave:

    • Thegetmark(),setmark() andgetmarkers() methods oftheWave_read andWave_write classeshave been deprecated since Python 3.13.

  • zipimport:

    • zipimport.zipimporter.load_module() has been deprecated sincePython 3.10. Useexec_module() instead.(gh-125746.)

Pending removal in Python 3.16

Pending removal in Python 3.17

  • collections.abc:

    • collections.abc.ByteString is scheduled for removal in Python 3.17.

      Useisinstance(obj,collections.abc.Buffer) to test ifobjimplements thebuffer protocol at runtime. For usein type annotations, either useBuffer or a unionthat explicitly specifies the types your code supports (e.g.,bytes|bytearray|memoryview).

      ByteString was originally intended to be an abstract class thatwould serve as a supertype of bothbytes andbytearray.However, since the ABC never had any methods, knowing that an object was aninstance ofByteString never actually told you anything usefulabout the object. Other common buffer types such asmemoryviewwere also never understood as subtypes ofByteString (either atruntime or by static type checkers).

      SeePEP 688 for more details.(Contributed by Shantanu Jain ingh-91896.)

  • typing:

    • Before Python 3.14, old-style unions were implemented using the private classtyping._UnionGenericAlias. This class is no longer needed for the implementation,but it has been retained for backward compatibility, with removal scheduled for Python3.17. Users should use documented introspection helpers liketyping.get_origin()andtyping.get_args() instead of relying on private implementation details.

    • typing.ByteString, deprecated since Python 3.9, is scheduled for removal inPython 3.17.

      Useisinstance(obj,collections.abc.Buffer) to test ifobjimplements thebuffer protocol at runtime. For usein type annotations, either useBuffer or a unionthat explicitly specifies the types your code supports (e.g.,bytes|bytearray|memoryview).

      ByteString was originally intended to be an abstract class thatwould serve as a supertype of bothbytes andbytearray.However, since the ABC never had any methods, knowing that an object was aninstance ofByteString never actually told you anything usefulabout the object. Other common buffer types such asmemoryviewwere also never understood as subtypes ofByteString (either atruntime or by static type checkers).

      SeePEP 688 for more details.(Contributed by Shantanu Jain ingh-91896.)

Pending removal in Python 3.19

  • ctypes:

    • Implicitly switching to the MSVC-compatible struct layout by setting_pack_ but not_layout_on non-Windows platforms.

  • hashlib:

    • In hash function constructors such asnew() or thedirect hash-named constructors such asmd5() andsha256(), their optional initial data parameter couldalso be passed a keyword argument nameddata= orstring= invarioushashlib implementations.

      Support for thestring keyword argument name is now deprecatedand slated for removal in Python 3.19.

      Before Python 3.13, thestring keyword parameter was not correctlysupported depending on the backend implementation of hash functions.Prefer passing the initial data as a positional argument for maximumbackwards compatibility.

Pending removal in Python 3.20

Pending removal in future versions

The following APIs will be removed in the future,although there is currently no date scheduled for their removal.

  • argparse:

    • Nesting argument groups and nesting mutually exclusivegroups are deprecated.

    • Passing the undocumented keyword argumentprefix_chars toadd_argument_group() is nowdeprecated.

    • Theargparse.FileType type converter is deprecated.

  • builtins:

    • Generators:throw(type,exc,tb) andathrow(type,exc,tb)signature is deprecated: usethrow(exc) andathrow(exc) instead,the single argument signature.

    • Currently Python accepts numeric literals immediately followed by keywords,for example0inx,1orx,0if1else2. It allows confusing andambiguous expressions like[0x1forxiny] (which can be interpreted as[0x1forxiny] or[0x1forxiny]). A syntax warning is raisedif the numeric literal is immediately followed by one of keywordsand,else,for,if,in,is andor. In a future release itwill be changed to a syntax error. (gh-87999)

    • Support for__index__() and__int__() method returning non-int type:these methods will be required to return an instance of a strict subclass ofint.

    • Support for__float__() method returning a strict subclass offloat: these methods will be required to return an instance offloat.

    • Support for__complex__() method returning a strict subclass ofcomplex: these methods will be required to return an instance ofcomplex.

    • Delegation ofint() to__trunc__() method.

    • Passing a complex number as thereal orimag argument in thecomplex() constructor is now deprecated; it should only be passedas a single positional argument.(Contributed by Serhiy Storchaka ingh-109218.)

  • calendar:calendar.January andcalendar.February constants aredeprecated and replaced bycalendar.JANUARY andcalendar.FEBRUARY.(Contributed by Prince Roshan ingh-103636.)

  • codecs: useopen() instead ofcodecs.open(). (gh-133038)

  • codeobject.co_lnotab: use thecodeobject.co_lines() methodinstead.

  • datetime:

    • utcnow():usedatetime.datetime.now(tz=datetime.UTC).

    • utcfromtimestamp():usedatetime.datetime.fromtimestamp(timestamp,tz=datetime.UTC).

  • gettext: Plural value must be an integer.

  • importlib:

    • cache_from_source()debug_override parameter isdeprecated: use theoptimization parameter instead.

  • importlib.metadata:

    • EntryPoints tuple interface.

    • ImplicitNone on return values.

  • logging: thewarn() method has been deprecatedsince Python 3.3, usewarning() instead.

  • mailbox: Use of StringIO input and text mode is deprecated, useBytesIO and binary mode instead.

  • os: Callingos.register_at_fork() in multi-threaded process.

  • pydoc.ErrorDuringImport: A tuple value forexc_info parameter isdeprecated, use an exception instance.

  • re: More strict rules are now applied for numerical group referencesand group names in regular expressions. Only sequence of ASCII digits is nowaccepted as a numerical reference. The group name in bytes patterns andreplacement strings can now only contain ASCII letters and digits andunderscore.(Contributed by Serhiy Storchaka ingh-91760.)

  • shutil:rmtree()’sonerror parameter is deprecated inPython 3.12; use theonexc parameter instead.

  • ssl options and protocols:

    • ssl.SSLContext without protocol argument is deprecated.

    • ssl.SSLContext:set_npn_protocols() andselected_npn_protocol() are deprecated: use ALPNinstead.

    • ssl.OP_NO_SSL* options

    • ssl.OP_NO_TLS* options

    • ssl.PROTOCOL_SSLv3

    • ssl.PROTOCOL_TLS

    • ssl.PROTOCOL_TLSv1

    • ssl.PROTOCOL_TLSv1_1

    • ssl.PROTOCOL_TLSv1_2

    • ssl.TLSVersion.SSLv3

    • ssl.TLSVersion.TLSv1

    • ssl.TLSVersion.TLSv1_1

  • threading methods:

  • typing.Text (gh-92332).

  • The internal classtyping._UnionGenericAlias is no longer used to implementtyping.Union. To preserve compatibility with users using this privateclass, a compatibility shim will be provided until at least Python 3.17. (Contributed byJelle Zijlstra ingh-105499.)

  • unittest.IsolatedAsyncioTestCase: it is deprecated to return a valuethat is notNone from a test case.

  • urllib.parse deprecated functions:urlparse() instead

    • splitattr()

    • splithost()

    • splitnport()

    • splitpasswd()

    • splitport()

    • splitquery()

    • splittag()

    • splittype()

    • splituser()

    • splitvalue()

    • to_bytes()

  • wsgiref:SimpleHandler.stdout.write() should not do partialwrites.

  • xml.etree.ElementTree: Testing the truth value of anElement is deprecated. In a future release itwill always returnTrue. Prefer explicitlen(elem) orelemisnotNone tests instead.

  • sys._clear_type_cache() is deprecated:usesys._clear_internal_caches() instead.

CPython bytecode changes

Pseudo-instructions

  • Add theANNOTATIONS_PLACEHOLDER pseudo instructionto support partially executed module-level annotations withdeferred evaluation of annotations.(Contributed by Jelle Zijlstra ingh-130907.)

  • Add theBINARY_OP_EXTEND pseudo instruction,which executes a pair of functions (guard and specialization functions)accessed from the inline cache.(Contributed by Irit Katriel ingh-100239.)

  • Add three specializations forCALL_KW;CALL_KW_PY for calls to Python functions,CALL_KW_BOUND_METHOD for calls to bound methods, andCALL_KW_NON_PY for all other calls.(Contributed by Mark Shannon ingh-118093.)

  • Add theJUMP_IF_TRUE andJUMP_IF_FALSE pseudo instructions,conditional jumps which do not impact the stack.Replaced by the sequenceCOPY1,TO_BOOL,POP_JUMP_IF_TRUE/FALSE.(Contributed by Irit Katriel ingh-124285.)

  • Add theLOAD_CONST_MORTAL pseudo instruction.(Contributed by Mark Shannon ingh-128685.)

  • Add theLOAD_CONST_IMMORTAL pseudo instruction,which does the same asLOAD_CONST, but is more efficientfor immortal objects.(Contributed by Mark Shannon ingh-125837.)

  • Add theNOT_TAKEN pseudo instruction, used bysys.monitoringto record branch events (such asBRANCH_LEFT).(Contributed by Mark Shannon ingh-122548.)

C API changes

Python configuration C API

Add aPyInitConfig C API to configure the Pythoninitialization without relying on C structures and the ability to makeABI-compatible changes in the future.

Complete thePEP 587PyConfig C API by addingPyInitConfig_AddModule() which can be used to add a built-in extensionmodule; a feature previously referred to as the “inittab”.

AddPyConfig_Get() andPyConfig_Set() functions to get and setthe current runtime configuration.

PEP 587 ‘Python Initialization Configuration’ unified all the waysto configure Python’s initialization. This PEP also unifies the configurationof Python’s preinitialization and initialization in a single API.Moreover, this PEP only provides a single choice to embed Python,instead of having two ‘Python’ and ‘Isolated’ choices (PEP 587),to further simplify the API.

The lower level PEP 587 PyConfig API remains available for use caseswith an intentionally higher level of coupling to CPython implementation details(such as emulating the full functionality of CPython’s CLI, including itsconfiguration mechanisms).

(Contributed by Victor Stinner ingh-107954.)

See also

PEP 741 andPEP 587

New features in the C API

Limited C API changes

Removed C APIs

  • Creatingimmutabletypes withmutable bases was deprecated in Python 3.12,and now raises aTypeError.(Contributed by Nikita Sobolev ingh-119775.)

  • RemovePyDictObject.ma_version_tag member, which was deprecatedin Python 3.12.Use thePyDict_AddWatcher() API instead.(Contributed by Sam Gross ingh-124296.)

  • Remove the private_Py_InitializeMain() function.It was aprovisional API added to Python 3.8 byPEP 587.(Contributed by Victor Stinner ingh-129033.)

  • Remove the undocumented APIsPy_C_RECURSION_LIMITandPyThreadState.c_recursion_remaining.These were added in 3.13 and have been removed without deprecation.UsePy_EnterRecursiveCall() to guard against runawayrecursion in C code.(Removed by Petr Viktorin ingh-133079, see alsogh-130396.)

Deprecated C APIs

Pending removal in Python 3.15

Pending removal in Python 3.16

  • The bundled copy oflibmpdec.

Pending removal in Python 3.18

Pending removal in future versions

The following APIs are deprecated and will be removed,although there is currently no date scheduled for their removal.

Build changes

  • PEP 776: Emscripten is now an officially supported platform attier 3. As a part of this effort, more than 25 bugs inEmscripten libc were fixed. Emscripten now includes supportforctypes,termios, andfcntl, as well asexperimental support for the newdefault interactive shell.(Contributed by R. Hood Chatham ingh-127146,gh-127683, andgh-136931.)

  • Official Android binary releases are now provided onpython.org.

  • GNU Autoconf 2.72 is now required to generateconfigure.(Contributed by Erlend Aasland ingh-115765.)

  • wasm32-unknown-emscripten is now aPEP 11 tier 3 platform.(Contributed by R. Hood Chatham ingh-127146,gh-127683, andgh-136931.)

  • #pragma-based linking withpython3*.lib can now be switched offwithPy_NO_LINK_LIB.(Contributed by Jean-Christophe Fillion-Robin ingh-82909.)

  • CPython now enables a set of recommended compiler options by defaultfor improved security.Use the--disable-safetyconfigure option to disable them,or the--enable-slower-safety option for a larger setof compiler options, albeit with a performance cost.

  • TheWITH_FREELISTS macro and--without-freelistsconfigureoption have been removed.

  • The newconfigure option--with-tail-call-interpmay be used to enable the experimental tail call interpreter.SeeA new type of interpreter for further details.

  • To disable the new remote debugging support, use the--without-remote-debugconfigure option.This may be useful for security reasons.

  • iOS and macOS apps can now be configured to redirectstdout andstderr content to the system log.(Contributed by Russell Keith-Magee ingh-127592.)

  • The iOS testbed is now able to stream test output while the test is running.The testbed can also be used to run the test suite of projects other thanCPython itself.(Contributed by Russell Keith-Magee ingh-127592.)

build-details.json

Installations of Python now contain a new file,build-details.json.This is a static JSON document containing build details for CPython,to allow for introspection without needing to run code.This is helpful for use-cases such as Python launchers, cross-compilation,and so on.

build-details.json must be installed in the platform-independentstandard library directory. This corresponds to the‘stdlib’sysconfig installation path,which can be found by runningsysconfig.get_path('stdlib').

See also

PEP 739build-details.json 1.0 – a static description filefor Python build details

Discontinuation of PGP signatures

PGP (Pretty Good Privacy) signatures will not be providedfor releases of Python 3.14 or future versions.To verify CPython artifacts, users must useSigstore verification materials.Releases have been signed usingSigstore since Python 3.11.

This change in release process was specified inPEP 761.

Free-threaded Python is officially supported

The free-threaded build of Python is now supported and no longer experimental.This is the start ofphase II wherefree-threaded Python is officially supported but still optional.

The free-threading team are confident that the project is on the right path,and appreciate the continued dedication from everyone working to makefree-threading ready for broader adoption across the Python community.

With these recommendations and the acceptance of this PEP, the Python developercommunity should broadly advertise that free-threading is a supportedPython build option now and into the future, and that it will not be removedwithout a proper deprecation schedule.

Any decision to transition tophase III,with free-threading as the default or sole build of Python is still undecided,and dependent on many factors both within CPython itself and the community.This decision is for the future.

Binary releases for the experimental just-in-time compiler

The official macOS and Windows release binaries now include anexperimentaljust-in-time (JIT) compiler. Although it isnot recommended for productionuse, it can be tested by settingPYTHON_JIT=1 as anenvironment variable. Downstream source builds and redistributors can use the--enable-experimental-jit=yes-off configuration option for similarbehavior.

The JIT is at an early stage and still in active development. As such, thetypical performance impact of enabling it can range from 10% slower to 20%faster, depending on workload. To aid in testing and evaluation, a set ofintrospection functions has been provided in thesys._jit namespace.sys._jit.is_available() can be used to determine if the current executablesupports JIT compilation, whilesys._jit.is_enabled() can be used to tellif JIT compilation has been enabled for the current process.

Currently, the most significant missing functionality is that native debuggersand profilers likegdb andperf are unable to unwind through JIT frames(Python debuggers and profilers, likepdb orprofile, continue towork without modification). Free-threaded builds do not support JIT compilation.

Please report any bugs or major performance regressions that you encounter!

See also

PEP 744

Porting to Python 3.14

This section lists previously described changes and other bugfixesthat may require changes to your code.

Changes in the Python API

  • On Unix platforms other than macOS,forkserver is now the defaultstart method formultiprocessingandProcessPoolExecutor, instead offork.

    See(1) and(2) for details.

    If you encounterNameErrors or pickling errors coming out ofmultiprocessing orconcurrent.futures, see theforkserver restrictions.

    This change does not affect Windows or macOS, where‘spawn’ remains the default start method.

  • functools.partial is now a method descriptor.Wrap it instaticmethod() if you want to preserve the old behavior.(Contributed by Serhiy Storchaka and Dominykas Grigonis ingh-121027.)

  • Thegarbage collector is now incremental,which means that the behavior ofgc.collect() changes slightly:

    • gc.collect(1): Performs an increment of garbage collection,rather than collecting generation 1.

    • Other calls togc.collect() are unchanged.

  • Thelocale.nl_langinfo() function now temporarily sets theLC_CTYPElocale in some cases.This temporary change affects other threads.(Contributed by Serhiy Storchaka ingh-69998.)

  • types.UnionType is now an alias fortyping.Union,causing changes in some behaviors.Seeabove for more details.(Contributed by Jelle Zijlstra ingh-105499.)

  • The runtime behavior of annotations has changed in various ways; seeabove for details. While most code that interactswith annotations should continue to work, some undocumented details may behavedifferently.

  • As part of making themimetypes CLI public,it now exits with1 on failure instead of0and2 on incorrect command-line parameters instead of1.Error messages are now printed to stderr.

  • The\B pattern in regular expression now matches the empty stringwhen given as the entire pattern, which may cause behavioural changes.

  • On FreeBSD,sys.platform no longer contains the major version number.

Changes in annotations (PEP 649 andPEP 749)

This section contains guidance on changes that may be needed to annotationsor Python code that interacts with or introspects annotations,due to the changes related todeferred evaluation of annotations.

In the majority of cases, working code from older versions of Pythonwill not require any changes.

Implications for annotated code

If you define annotations in your code (for example, for use with a static typechecker), then this change probably does not affect you: you can keepwriting annotations the same way you did with previous versions of Python.

You will likely be able to remove quoted strings in annotations, which are frequentlyused for forward references. Similarly, if you usefrom__future__importannotationsto avoid having to write strings in annotations, you may well be able toremove that import once you support only Python 3.14 and newer.However, if you rely on third-party libraries that read annotations,those libraries may need changes to support unquoted annotations before theywork as expected.

Implications for readers of__annotations__

If your code reads the__annotations__ attribute on objects,you may want to make changes in order to support code that relies ondeferred evaluation of annotations.For example, you may want to useannotationlib.get_annotations() withtheFORWARDREF format,as thedataclasses module now does.

The externaltyping_extensions package provides partial backportsof some of the functionality of theannotationlib module,such as theFormat enum andtheget_annotations() function.These can be used to write cross-version code that takes advantage ofthe new behavior in Python 3.14.

Related changes

The changes in Python 3.14 are designed to rework how__annotations__works at runtime while minimizing breakage to code that containsannotations in source code and to code that reads__annotations__.However, if you rely on undocumented details of the annotation behavioror on private functions in the standard library, there are many ways in whichyour code may not work in Python 3.14.To safeguard your code against future changes, only use the documentedfunctionality of theannotationlib module.

In particular, do not read annotations directly from the namespace dictionaryattribute of type objects.Useannotationlib.get_annotate_from_class_namespace() during classconstruction andannotationlib.get_annotations() afterwards.

In previous releases, it was sometimes possible to access class annotationsfrom an instance of an annotated class. This behavior was undocumentedand accidental, and will no longer work in Python 3.14.

from__future__importannotations

In Python 3.7,PEP 563 introduced thefrom__future__importannotationsfuture statement, which turns all annotations into strings.

However, this statement is now deprecated and it is expected to be removedin a future version of Python.This removal will not happen until after Python 3.13 reaches its end of lifein 2029, being the last version of Python without support for deferredevaluation of annotations.

In Python 3.14, the behavior of code usingfrom__future__importannotationsis unchanged.

Changes in the C API

  • Py_Finalize() now deletes all interned strings. Thisis backwards incompatible to any C extension that holds onto an internedstring after a call toPy_Finalize() and is then reused after acall toPy_Initialize(). Any issues arising from this behavior willnormally result in crashes during the execution of the subsequent call toPy_Initialize() from accessing uninitialized memory. To fix, usean address sanitizer to identify any use-after-free coming froman interned string and deallocate it during module shutdown.(Contributed by Eddie Elizondo ingh-113601.)

  • TheUnicode Exception Objects C APInow raises aTypeError if its exception argument is notaUnicodeError object.(Contributed by Bénédikt Tran ingh-127691.)