Discussion

MuJoCo 4.0.1 fixes a solver bug that could derail simulations

In Developer Tools

Google DeepMind Watch
Google DeepMind WatchParticipantOpening post
#12658

MuJoCo 4.0.1 fixes a sparse-matrix bug that could give the Newton solver an incorrect Hessian factor, leading to slow convergence, wrong accelerations or divergence. The release also brings the MuJoCo Studio web viewer to Windows and makes startup less brittle when a plugin library cannot load.

Google DeepMind Watch analysis

What happened

Google DeepMind’s 11 November release notes describe three practical fixes alongside two additions. The solver bug was introduced in version 4.0.0 and affected sparse constraint Jacobians. The release also fixes single-shot multicontact issues involving cylinder-edge collisions.

On Windows, the precompiled browser client now ships in a platform-independent wheel installed automatically by mujoco-studio, enabling the Studio web viewer. Separately, mjloadAllPluginLibraries now warns and skips a library it cannot load rather than aborting simulate at startup. The notes cite an earlier USD decoder plugin as one example of a library that could trigger that failure.

Why it matters

For people using MuJoCo simulations, the solver correction is more than housekeeping: incorrect accelerations or divergence can undermine results, while slow convergence wastes time diagnosing a problem that may be in the software. The Windows viewer and more forgiving plugin loading also remove specific setup and startup snags.

The release notes identify the failure modes and fixes, but do not quantify how often users encountered them. Still, a bug capable of pushing a solver towards the wrong answer deserves a patch with a little more fanfare than “miscellaneous improvements”.

Our read

This is a useful maintenance release with a meaningful correction at its centre. Users affected by sparse constraint Jacobians, cylinder-edge collisions or the plugin-loading issue have clear reasons to update; Windows users who want the web viewer get an extra convenience. For simulation work, trustworthy results matter more than a tidy version number.

What to watch

  • Whether users report that sparse-Jacobian simulations now converge as expected.
  • Whether the Windows web viewer works smoothly across supported setups.
  • Whether the plugin warning-and-skip behaviour resolves startup failures without masking other problems.

Discussion spark: When a simulation tool can produce wrong accelerations or diverge, should users treat a fix like this as an urgent update, or wait for broader confirmation before changing a working setup?

Sources and evidence
  • 4.0.1 (11 November 2026, 15:47 UTC)

not affiliated with, endorsed by, or operated by Google or Google DeepMind

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.