Where to Find the Release
You can get the latest version of the Intel® Distribution for Python from the Intel channel. The installation guide for Conda, Mamba, and PIP is available here.
|
Release version |
Release Date |
|
2026.0 |
April 28, 2026 |
New in this Release
2026.0
Highlights
Intel® Data Parallel Control Library:
The following changes were made to the dpctl.tensor submodule (that is now dpnp.tensor) since the last release:
- Updated dlpack version to 1.2
- Improved performance of conversions from boolean arrays
- Disallowed conversion of any arrays with dimensions (including those of size-1) to Python scalars, following NumPy
Intel® Data Parallel Extension for NumPy*:
- Added implementation of `dpnp.linalg.lu_solve` for batch inputs (SciPy-compatible)
- Added `dpnp.exceptions` submodule to aggregate the generic exceptions used by dpnp
- Added implementation of `dpnp.scipy.special.erfcx`
- Added implementation of `dpnp.scipy.special.erfinv` and `dpnp.scipy.special.erfcinv`
- Added implementation of `dpnp.ndarray.tolist` method
- Added implementation of `dpnp.frexp`
- Added implementation of `dpnp.ndarray.tofile` method
- Added implementation of `dpnp.ndarray.tobytes` method
- Added implementation of `dpnp.divmod`
- Added implementation of `dpnp.isin` function
- Added implementation of `dpnp.scipy.linalg.lu` (SciPy-compatible)
- Added support for ndarray subclassing via `dpnp.ndarray.view` method with `type` parameter
Improvements and changes
Intel® Data Parallel Extension for NumPy*:
- Changed the license from `BSD-2-Clause` to `BSD-3-Clause`
- Redesigned `dpnp.modf` function to be a part of `ufunc` and `vm` pybind11 extensions
- Added support for the `out` keyword to accept a tuple, bringing ufunc signatures into alignment with those in NumPy
- Unified public API definitions in `dpnp.linalg` and `dpnp.scipy` submodules
- Aligned the signature of `dpnp.reshape` function with Python array API by making `shape` a required argument
- Aligned `dpnp.trim_zeros` with NumPy 2.4 to support a tuple of integers passed with `axis` keyword
- Aligned `strides` property of `dpnp.ndarray` with NumPy and CuPy implementations
- Extended `dpnp.nan_to_num` to support broadcasting of `nan`, `posinf`, and `neginf` keywords
- Changed `dpnp.partition` implementation to reuse `dpnp.sort` which brings the performance benefit
Fixed issues
- Resolved an issue causing `dpnp.linspace` to return an incorrect output shape when inputs were passed as arrays
- Resolved an issue where `dpnp` always returns the base allocation pointer, when the view start is expected
- Fixed an issue causing an exception in `dpnp.geomspace` and `dpnp.logspace` when called with explicit `device` keyword but any input array is allocated on another device
- Fixed `.data.ptr` property on array views to correctly return the pointer to the view's data location instead of the base allocation pointer
- Resolved an issue with strides calculation in `dpnp.diagonal` to return correct values for empty diagonals
- Ensured device aware dtype handling in `dpnp.identity` and `dpnp.gradient`
Known Issues and Limitations
No critical issues have been identified at this time. For platform-specific considerations and limitations, please refer to the documentation. Report any issues via the Intel® Distribution for Python Community Forum.
System Requirements
Supported Architectures and Terminology - Intel® 64 Architecture refers to systems based on IA-32 architecture processors which have 64-bit architectural extensions (like Intel® Core™ architecture processors, running a 64-bit operating system such as Microsoft Windows* 10 x64 or a Linux* "x86_64" variant).
Note: Minor releases (e.g., 2026.1) automatically inherit all OS requirements from the base major release (2026.0).Only operating systems marked with (+) are newly added in that minor release, and those marked with (–) are removed. Asterisks(**) indicates Deprecations. All others remain unchanged and are not duplicated in the table.
Hardware Requirements
|
Release Version |
Supported CPUs |
Supported GPUs |
|
2026.0 (Baseline) |
|
|
Supported Operating Systems
- Linux*
- Ubuntu* 22.04 & higher (CPU & GPU)
- SUSE Linux Enterprise Server* 15 SP4 (CPU & GPU), 15 SP5 (CPU & GPU), 15 SP6 (CPU & GPU), 15 SP7 (CPU & GPU), 16.0 (CPU)
- Red Hat* Enterprise Linux 8, 9, 10 (CPU & GPU)
- Windows*
- Windows Pro & Enterprise 10, 11
- Windows Server 2019* (CPU & GPU), 2022* (CPU & GPU), 2025* (CPU)
- Fedora 41, 42 (CPU)
- Debian 11, 12 (CPU)
Software Requirements
|
Release Version |
Supported versions |
Package managers |
Compatible IDEs |
|
2026.0 |
|
|
|
Graphics Driver Installation
Linux
The driver packages needed on Linux are described in the Data Center GPU Series Driver Installation and Installing Client GPUs pages.
Windows
For GPU development, the latest GPU drivers need to be installed. They can be downloaded at Intel® Arc™ & Iris® Xe Graphics - Windows*.
Support Deprecated
Intel® Data Parallel Extension for NumPy*:
- `dpnp.asfarray` is deprecated. Use `dpnp.asarray` with an appropriate dtype instead
- Passing the output array ``out`` positionally to `dpnp.minimum` and `dpnp.maximum` is deprecated. Pass the output with the keyword form, e.g. ``dpnp.minimum(a, b, out=c)``
- `dpnp.ndarray.T` property is deprecated for non two-dimensional array to be compatible with the Python array API standard. To achieve a similar behavior when ``a.ndim != 2``, either ``a.transpose()``, or ``a.mT`` (swaps the last two axes only), or ``dpnp.permute_dims(a, range(a.ndim)[::-1])`` can be used
- `dpnp.fix` is deprecated. Use `dpnp.trunc` instead, which provides identical functionality
Support Removed
Intel® Data Parallel Extension for NumPy*:
- Dropped support for Python 3.9
- Removed the `newshape` parameter from `dpnp.reshape`, which has been deprecated since dpnp 0.17.0. Pass it positionally or use `shape=` on newer versions
- Dropped a conda run dependency on `onemkl-sycl-stats` package
Intel® Data Parallel Control Library:
- Removed previously-deprecated dpctl.tensor submodule. It is now part of Data Parallel Extensions for Numpy* and can be used identically by importing dpnp.tensor in place of dpctl.tensor.
Other Documentation and Support
You can visit the technical support forum, and other support information at:
Attributions
The license information for third party software that was used to develop the Intel® Distribution for Python can be found in <install_dir>/licensing/third-party-programs.txt .
For the avoidance of doubt, the Intel® Distribution for Python is solely governed by the terms and conditions of the End User License Agreement for Intel® Software Development Product that accompaniesthe Intel® Distribution for Python.
Legal Information
No license (express or implied, by estoppel or otherwise) to any intellectual property rights is granted by this document.
Intel disclaims all express and implied warranties, including without limitation, the implied warranties of merchantability, fitness for a particular purpose, and non-infringement, as well as any warranty arising from course of performance, course of dealing, or usage in trade.
This document contains information on products, services and/or processes in development. All information provided here is subject to change without notice. Contact your Intel representative to obtain the latest forecast, schedule, specifications and roadmaps.
The products and services described may contain defects or errors which may cause deviations from published specifications. Current characterized errata are available on request. No product or component can be absolutely secure.
Copies of documents which have an order number and are referenced in this document may be obtained by calling 1-800-548-4725 or by visiting Resource & Documentation Center.
Intel, the Intel logo, Intel Core, Intel Xeon Phi, VTune, and Xeon are trademarks of Intel Corporation in the U.S. and/or other countries.
*Other names and brands may be claimed as the property of others.
Microsoft, Windows, and the Windows logo are trademarks, or registered trademarks of Microsoft Corporation in the United States and/or other countries.
Copyright 2026 Intel Corporation.
This software and the related documents are Intel copyrighted materials, and your use of them is governed by the express license under which they were provided to you (License). Unless the License provides otherwise, you may not use, modify, copy, publish, distribute, disclose or transmit this software or the related documents without Intel's prior written permission.
This software and the related documents are provided as is, with no express or implied warranties, other than those that are expressly stated in the License.