| Project | Release | Key Features | Why It Matters | |---------|---------|--------------|----------------| | (final maintenance release) | 3.13.2 (Nov 4) | • Faster CPython start‑up (≈ 7 % on average) • Full support for PEP 701 – Pattern Matching Enhancements • Deprecation of the distutils module (complete removal slated for 3.14) • New typing.Final runtime checks | Consolidates the pattern‑matching syntax introduced in 3.10, making it a first‑class tool for data‑centric code. The start‑up speed win is a boon for serverless and CLI workloads. | | NumPy 2.0 | 2.0.0 (Nov 12) | • Native support for structured tensors (named axes) • SIMD‑accelerated reductions on ARM Neoverse • Drop of Python 3.8 support | The “named axes” API brings NumPy a step closer to the ergonomics of Xarray while staying lightweight. The ARM improvements are a direct response to the surge in edge‑AI devices. | | pandas 2.2 | 2.2.0 (Nov 18) | • Lazy‑evaluation engine (optional) • Built‑in polars ‑compatible back‑end • Enhanced JSON handling with streaming support | Lazy evaluation lets massive DataFrames be manipulated with near‑zero memory overhead, a game‑changer for ETL pipelines that previously relied on Dask. | | PyTorch 2.5 | 2.5.0 (Nov 22) | • Python 3.13‑only wheels (no legacy support) • torch.compile now supports dynamic shape tracing • Integrated torchserve CLI for quick model deployment | The tighter coupling with Python 3.13 encourages the community to upgrade, while dynamic tracing eases production use of variable‑length inputs (e.g., NLP). | | FastAPI 0.110 | 0.110.1 (Nov 27) | • Automatic OpenAPI 3.1 generation • Built‑in support for async‑first ORMs (SQLModel 2.0) • Deprecation of starlette ‑only middleware | The OpenAPI upgrade aligns FastAPI with the latest specification, improving compatibility with API‑gateways and client generators. |

November 2025 also served as a stark reminder of the language's evolution as .