●Change8

Migrating to JAX jax-v0.11.0

Version jax-v0.11.0 introduces 4 breaking changes. This guide details how to update your code.

Released: 7/16/2026

4
Breaking Changes
5
Migration Steps
20
Affected Symbols

⚠️ Check Your Code

If you use any of these symbols, you need to read this guide:

jax.experimental.hijaxjax.experimental.hijax.linearize_from_jvpjax.experimental.hijax.apply_derived_linearizationjax.experimental.hijax.vjp_fwd_from_jvpjax.experimental.hijax.transpose_jvpjax.experimental.hijax.vjp_fwd_from_linjax.experimental.hijax.transpose_linearizedjax.experimental.hijax.jvp_from_linjax.custom_rematjax.checkpoint_policiesjax.checkpoint_policies.SaveOnlyTheseNamesjax.checkpoint_policies.SaveAnyNamesButThesejax.checkpoint_policies.SaveAndOffloadOnlyTheseNamesjax.Inlinejax.jitjax.numpy.emptyjax.numpy.empty_likejax.numpy.crossjax.corejax.interpreters.pxla

Breaking Changes

●Issue #1

The deprecated module jax.cloud_tpu_init was removed. References to it can be safely removed.

●Issue #2

Support for Python 3.11, NumPy 2.0, and SciPy 1.14 has been dropped. Update your Python, NumPy, and SciPy versions.

●Issue #3

Support for Python 3.13 free-threaded (`3.13t`) has been dropped. Use Python 3.14 free-threaded or a stable Python 3.13 build.

●Issue #4

`jax.numpy.empty` and `jax.numpy.empty_like` now produce uninitialized arrays. To initialize arrays to zeros, use `jax.numpy.zeros` or `jax.numpy.zeros_like` instead.

Migration Steps

  1. 1
    Update your Python version to 3.12 or later.
  2. 2
    Update your NumPy version to 2.0 or later.
  3. 3
    Update your SciPy version to 1.14 or later.
  4. 4
    If you were using `jax.cloud_tpu_init`, remove all references to it.
  5. 5
    If you need the previous behavior of `jax.numpy.empty` and `jax.numpy.empty_like` (initializing to zeros), switch to `jax.numpy.zeros` and `jax.numpy.zeros_like` respectively.

Release Summary

This release introduces experimental hijax API for custom derivative rules and `jax.custom_remat` for per-function rematerialization control. It also drops support for older Python, NumPy, and SciPy versions and modifies the behavior of `jax.numpy.empty`.

Need More Details?

View the full release notes and all changes for JAX jax-v0.11.0.

View Full Changelog