The Python ecosystem for IBM Power continues its rapid evolution. Today, we're excited to announce a major milestone: Python 3.14 support is now available, alongside significant package additions and enhanced security transparency with CVE data that strengthens the platform's capabilities for modern AI, machine learning and data analytics workloads.
Python 3.14: Future-Ready Performance
Python 3.14 brings enhanced performance optimizations and language features that developers have been eagerly anticipating. With this release, IBM Power users can leverage these improvements immediately, maintaining parity with the broader Python ecosystem while benefiting from Power architecture's computational advantages.
70+ packages now support Python 3.14, including critical packages across the stack:
- Deep Learning: PyTorch, TorchVision, TorchAudio
- Scientific Computing: NumPy, SciPy, Pandas, Matplotlib, Statsmodels
- Performance: Numba, llvmlite, TBB, Cython
- Computer Vision: OpenCV, Pillow, tesserocr
- NLP & Text: tiktoken, regex, lxml, MarkupSafe
- Data Processing: PyArrow, H5py, array_record, fastavro, sentencepiece
- Database: PyMongo, PyODBC, Snowflake Connector, Cassandra Driver
- Web & Networking: cryptography, greenlet, httptools, multidict, yarl
- System Libraries: zstandard, librt, xattr
60+ Brand New Packages Across All Python Versions
This release introduces 60+ new packages available for Python 3.9 through 3.14, significantly expanding capabilities across multiple domains:
- ML & AI: CatBoost, Spacy, Thinc, Blis, Cymem, XGrammar, Jaxlib, Einops, Gradio
- Data & Analytics: Polars, FastAPI, FSSpec, Pydantic, Boto3, S3Transfer
- LLM Frameworks: LangChain-classic, LangFlow, LangGraph, OpenAI-harmony, Tiktoken
- Infrastructure: Ansible-navigator, AWX, nodejs-wheel-binaries, Watchfiles, Pyro4
- Database: Snowflake-connector-python, sqlite-vec, PyCurl
- Security & Dev Tools: PyNaCl, Truststore, SELinux, Pytest-mock, Pygments, Sphinx extensions, Cython, Cforge
Featured Packages with New Examples
CatBoost: Gradient Boosting Excellence
- CatBoost 1.2.7 brings Yandex's gradient boosting library to IBM Power with native categorical feature handling.
- Use Case: Financial fraud detection, customer churn prediction, recommendation systems, patient outcome prediction.
Polars: Lightning-Fast DataFrames
- Polars 1.38.1 delivers high-performance DataFrames with Rust-based query optimization and parallel execution.
- Use Cases: Large-scale ETL pipelines, processing datasets larger than RAM, real-time market data analysis, genomics data processing.
Tiktoken: Fast LLM Tokenization
- Tiktoken provides OpenAI's fast tokenizer for efficient GPT model token processing.
- Use Cases: LLM applications, chatbots, text preprocessing, token counting, cost estimation.
Examples: Production-ready examples for CatBoost, Polars, and Tiktoken are available in the pyeco repository.
Release Statistics
- 180+ packages in this release (60+ new packages and 120+ updated packages)
- 70+ packages with Python 3.14 support
- 60+ new packages across Python 3.9 - 3.14
- 120+ package updates with latest versions
- 340+ unique packages total
- 3,800+ wheels in the repository
Enhanced Security and Transparency
The Package Index now includes CVE (Common Vulnerabilities and Exposures) data alongside license information for all wheels, enabling informed security decisions:
Refer to
DevpiWheelsIndex.md for detailed information on each wheel, including license data and associated CVE IDs. A sample entry is provided below for reference:
|
Wheel File
|
License
|
CVE IDs
|
|
black-22.12.0+ppc64le1-py3-none-any.whl
|
MIT, Python-2.0
|
GHSA-3936-cmfr-pm3m,
GHSA-fj7x-q9j7-g6q6
|
This transparency helps you to:
- Assess license compliance requirements.
- Identify known vulnerabilities (CVE IDs).
- Make informed security decisions for your deployments.
Getting Started
Install packages using:
pip install <package-name> --prefer-binary --extra-index-url=https://wheels.developerfirst.ibm.com/ppc64le/linux
Examples:
pip install catboost --prefer-binary --extra-index-url=https://wheels.developerfirst.ibm.com/ppc64le/linux
pip install polars --prefer-binary --extra-index-url=https://wheels.developerfirst.ibm.com/ppc64le/linux
Examples
The pyeco repository contains production-ready examples:
- CatBoost, Polars, Tiktoken: New package examples
- PyTorch, LightGBM, XGBoost: Deep learning and gradient boosting
- ML-DS Pipeline: 20+ library workflows
- More: Azure, Langchain, ONNX, Ray, vLLM
Each example includes installation scripts, requirements, and test suites.
What's Next
We're actively expanding package coverage, optimizing performance for IBM Power systems.
If a required Python package or specific version is not available as an IBM Power–optimized wheel, IBM provides a formal enablement process through the ISV Enablement Program.
The future of Python on IBM Power is here. Start building today.
Resources