Installing Mathics3

Here we describe the various ways to get Mathics3 installed.

Mathics3 (mathics-core) Requirements

The Mathics3 suite runs on Python 3.7 or later. We also support Pyston and PyPy 3.7 or later.

Underneath, Mathics3 core relies on SymPy which relies on NumPy. These and the other requirements will be installed automatically if you use the standard Python installer pip. They are also listed in

SciPy is optional. It is used for images and provides alternative implementations for a number of builtins.

Building the PDF documentation has a number of additional dependencies.

  • xetex 3.14159265-2.6-0.999991 or greater

  • asymptote, 2.83 or greater

  • ghostscript Version 9.56.1 or greater

Pay close attention to the version. Some OS-provided packages have bugs in them that will prevent certain images like those involving opacity (used in Filling) not to render.

OS package dependencies

Here we will only describe OS package dependencies for Mathics3. For the front-ends, Mathics-Django or mathicsscript, see :ref`Install from the Mathics3 Github Organization` for specific github repositories.


For Debian/Ubuntu based systems:

$ apt-get install python-dev libsqlite3-dev python-setuptools liblapack-dev llvm-dev

as super-user, i.e. either after having issued su or by preceding the command with sudo).

Note that to build the PDF, you will need asymptote and .deb package has a couple of bugs that will cause some graphs with opacity to fail. Asymptote version 2.71 or later is recommended. I build this from the git source.

Also ghostscript is needed and 9.50 has bugs in it too that will cause failures in rendering Asymptote images. Use 9.54 or later.


On Mac OS X:

$ brew install sqlite3
$ brew install llvm@11

Make sure when you install llvmlite to specify the LLVM_CONFIG:

$ LLVM_CONFIG=/usr/local/Cellar/llvm@11/11.1.0/bin/llvm-config pip install llvmlite


On FreeBSD:

$ sudo pkg install math/py-mathics

MS Windows

On Microsoft Windows:

$ choco install llvm

Install from PyPI

If you have the appropriate OS-dependent packages installed, and want to install everything in one shot, try:

$ pip install Mathics-omnibus[full]

This might not work if you don’t have the full suite of support OS packages LLVM, xetex and numerous others, the above may fail.

You can try a more minimal installation using:

$ pip install Mathics-omnibus

Important Note: when installing the full suite of programs as is done in Mathics-omnibus, all of the components are guarenteed to work either if everything is installed via PyPI or installed using current sources. Mixing current sources with packages from PyPI may not work because as of 2023 and for a little while longer our API keeps changing a bit drastically. At some point in the far future the API will become more stable.

When you install Mathics-omnibus you are this suite of independent Mathics3 packages:

These packages in turn pull in two other Mathics3 packages:

and possibly:

Above, the name in parenthesis is the Python import module name you would use to import from that package.

The Mathics-omnibus Python package doesn’t have any code per se in it. Instead, it just contains dependencies to other PyPI Mathics3 packages.of other PyPI package.

It is more or less equivalent to:

$ pip install Mathics-Scanner # Mathics3 scanner and builtin operator definitions
$ pip install Mathics3  # this is the core engine. It is a dependency of some of the below too
$ pip install Mathics-Django[full] # web front-end with extras
$ pip install mathicsscript[full]  # the command-line interface with extras
$ pip install pymathics-natlang # the Natural-language Python module
$ pip install pymathics-graph # the Python module for working with Graphs and Networks

If something fails, try using the above commands one by one, and remove the [full] to get a more basic installation.

Note the name “Mathics3” for the core engine. This is the most recent release. “Mathics” has pre-Python 3 code.

Other sections have descriptions of the various pieces, what features they add and what dependencies are needed to run them.

If you want mathics-core to include modules that have been run through Cython, then install Cython separately:

$ pip install cython

From docker (dockerhub)

As an alternative to building all the components from source or via Python package, you can run pre-built code for all of the components via docker. To download a copy of the docker image run:

$ docker pull mathicsorg/mathics

This will pull the latest development release that has a docker tag name #latest if you want a stable release, give the version in the pull command. For example

$ docker pull mathicsorg/mathics#5.0.0

From an OS-Specific Repository

Click on the link below from for details for a specific OS and distribution.

Packaging status

Install from the Mathics3 Github Organization

If you need the most recent fixes, you can install from github. Note that there isn’t a single github repository that contains everything.

Also, since Mathics3 is in constant flux, dependencies can change. The development version from might require development version from another.

The minimal set of Mathics3 Python packages that need to be installed is:

When Mathics3 (which depends on Mathics-Scanner) is installed, there is a minimalist command-line utility called mathics available which allows you to enter Mathics3 statements. For help on this type:

mathics --help

There are more filled-out front ends. mathicsscript is a more full featured command-line script similar to wolframscript. There is a Django-based front-end called mathics-django.

There are also two Mathics3 modules written in Python:

For each of the packages above installing is about the same:

$ git clone <name-of-repository>
$ cd <name-of-repository>
$ pip install -e .   # or pip install -e .[full]
$ make check # to test code

To run the Django-based front-end type:

$ make runserver