Set up your bench: install, first run, and where to look

The steps, in full

1. Check Python. QuTiP’s package metadata requires Python 3.11 or later; scqubits’ README lists Python 3.10 to 3.12 (its metadata says 3.9 or later). Python 3.11 or 3.12 satisfies both; this part’s examples were run on Python 3.13.7, which both packages installed on.

   python --version

2. Make a private environment so the two packages and their dependencies stay out of your system Python. On Windows (PowerShell):

   python -m venv bench
   bench\Scripts\Activate.ps1

   On macOS or Linux:

   python3 -m venv bench
   source bench/bin/activate

   The prompt now starts with (bench). QuTiP’s own instructions describe the conda route instead (conda create -n qutip-env python, then conda install qutip after adding the conda-forge channel with the lowest priority); scqubits’ page advises against mixing pip into a conda environment.
3. Install. QuTiP and scqubits are on PyPI. Pin the versions below. With the newest QuTiP (5.3.1 when this was written) scqubits 4.3.1 can diagonalise a single qubit but fails as soon as you combine two systems in a HilbertSpace (see the table below), so this part runs on QuTiP 5.2.0 with SciPy 1.14.1. Add qutip-qip for named gates.

   pip install "qutip==5.2.0" "scqubits==4.3.1" "scipy==1.14.1" qutip-qip

   pip install "scqubits[gui]" jupyterlab

   The first line is the environment every example in this part was run in. The second adds optional packages (the gui extra of scqubits brings ipywidgets, ipyvuetify and h5py for its notebook widgets, and JupyterLab runs notebooks); we did not use them here.
4. Check both. Each package has an about() that prints its version and the versions of what it was built on.

   python -c "import qutip; qutip.about()"
   python -c "import scqubits; scqubits.about()"

   QuTiP Version:      5.2.0
   Numpy Version:      2.2.6
   Scipy Version:      1.14.1
   Matplotlib Version: 3.11.2
   Python Version:     3.13.7
   ...
   scqubits version: 4.3.1
   numpy version:    2.2.6
   scipy version:    1.14.1
   QuTiP version:    5.2.0

5. Open a notebook. A notebook runs code in cells and shows plots under them. With the environment active, jupyter lab opens one in your browser; VS Code and Google Colab also run notebooks (in Colab, put !pip install qutip scqubits in the first cell). Run a cell with Shift and Enter. Type a function name followed by a question mark, for example qt.mesolve?, to read its documentation without leaving the notebook.
6. Run the first line of real physics.

   import qutip as qt
   print(qt.sigmaz())

   Quantum object: dims=[[2], [2]], shape=(2, 2), type='oper', dtype=CSR, isherm=True
   Qobj data =
   [[ 1.  0.]
    [ 0. -1.]]

   If that prints, the next topic starts from here.

When it goes wrong. The messages below are the ones we met while writing this part, copied from real runs.

Where each tool’s own pages are. QuTiP’s installation page, checked on 25 October 2026, lists Python 3.11 or later, NumPy 1.23.2 or later and SciPy 1.13 or later, and optional Matplotlib for plots and Cython for compiling time-dependent terms. The qutip-tutorials site lists these sections: Python Introduction, Visualization, Quantum information processing, Time evolution, Optimal control, Tomography, Permutational invariant Lindblad dynamics, Hierarchical Equations of Motion, Miscellaneous tutorials and a lecture series on quantum mechanics with QuTiP. The scqubits-examples repository’s examples folder holds, among others, demo_transmon.ipynb, demo_fluxonium.ipynb, demo_hilbertspace.ipynb, demo_parametersweep.ipynb, demo_explorer.ipynb, demo_noise.ipynb, demo_gui.ipynb, demo_customcircuit.ipynb, demo_jaynes-cummings.ipynb and demo_qutip_fluxoniumcz.ipynb. Documentation sites move; if a link here has gone, search the tool’s name with the topic.