Python 3.14.6 is out with 179 bugfixes, build improvements, and documentation changes since 3.14.5. But the real story is the feature set that 3.14 introduced, which continues to reshape how Python developers write code.
Template String Literals (t-strings)
PEP 750 introduces t-strings, a new string type that works like f-strings but returns a tree structure instead of a plain string. This lets you build custom processors that handle interpolation safely, which is especially useful for generating SQL queries, HTML templates, or any context where automatic string interpolation is a security risk.
Where f-strings evaluate expressions and embed them directly, t-strings give you the parsed structure so you can escape, transform, or validate each piece before rendering.
Free-Threaded Python
PEP 779 makes free-threaded Python officially supported. This removes the Global Interpreter Lock (GIL) as a compile-time option, allowing true multi-threaded parallelism for CPU-bound Python code. It has been experimental since 3.13, and 3.14 marks the first release where it ships as a supported configuration.
Deferred Annotations (PEP 649)
Python now defers the evaluation of type annotations until they are actually accessed. Previously, annotations were evaluated at class definition time, which caused issues with forward references and circular imports. With PEP 649, you can annotate functions and classes with types that have not been defined yet, and the interpreter handles the resolution lazily.
Zstandard Compression in the Standard Library
PEP 784 adds the compression.zstd module, giving Python built-in support for the Zstandard compression algorithm. Zstandard offers better compression ratios and faster decompression than zlib, making it a strong default for data serialization, log compression, and network protocols. No more pip install python-zstandard for basic use cases.
Other Notable Changes
- PEP 758: except and except* expressions can now omit parentheses (cleaner exception handling syntax)
- PEP 768: Zero-overhead external debugger interface for CPython, enabling tools like debugpy to attach without performance penalties
- PEP 734: Multiple interpreters in the standard library for subinterpreter support
- JIT compiler: Official macOS and Windows binaries now include an experimental JIT compiler
- Android binaries: Official Android binary releases are now available for the first time
- HMAC: Built-in HMAC implementation with formally verified code from the HACL* project
- PyREPL: Syntax highlighting in the interactive Python REPL
- UUID: Versions 6-8 now supported, with versions 3-5 generation up to 40% faster
Build Changes
Python 3.14 drops PGP signatures for release artifacts (PEP 761), recommending Sigstore instead. This is a significant shift in the Python release security model, moving away from the aging PGP web-of-trust model toward Sigstore transparency-log-based verification.
Frequently Asked Questions
What are t-strings in Python 3.14?
T-strings (template string literals) are a new string type introduced in PEP 750. They work like f-strings but return a structured tree instead of a flat string, allowing you to safely process interpolated values before rendering them. This is useful for preventing injection attacks in SQL, HTML, and other contexts.
Is free-threaded Python ready for production?
Free-threaded Python is officially supported in 3.14 but is still marked as experimental. Most pure-Python code works without changes, but C extensions that depend on the GIL may need updates. Test your workload thoroughly before deploying in production.
How do I install Python 3.14.6?
Download from python.org/downloads, or use your system package manager. On Ubuntu, use the deadsnakes PPA. On macOS, use Homebrew (brew install python@3.14). On Windows, download the installer from python.org.
What is the Zstandard module for?
The new compression.zstd module provides built-in Zstandard compression and decompression. Zstandard is widely used in databases (RocksDB, PostgreSQL), package managers, and data pipelines for its speed and compression ratio. Having it in the standard library eliminates a common third-party dependency.
Does the JIT compiler improve performance?
The experimental JIT compiler included in 3.14 official binaries can improve performance for certain workloads, particularly loops and repetitive computation. It is opt-in and requires building from source for full control, but the binaries ship with it enabled by default.
