diff --git a/docs/pacemaker/install.md b/docs/pacemaker/install.md
index 8875d55..e359a3f 100644
--- a/docs/pacemaker/install.md
+++ b/docs/pacemaker/install.md
@@ -9,16 +9,26 @@ mamba activate ace
 
 You can use `conda` instead of `mamba`.
 
-## Install pacemaker and the GRACE backend
+## Install pacemaker
 
 ```bash
 python -m pip install --upgrade pip
-python -m pip install 'python-ace==0.4.0rc1' 'tensorpotential==0.6.2.1' 'numpy>=2,<2.4' 'tensorflow~=2.20.0' 'tf-keras~=2.20.0'
+python -m pip install 'python-ace==0.4.0'
 ```
 
-The GRACE backend is published on PyPI as `tensorpotential`.
-Use the versions above together; this backend release requires NumPy below
-2.4 for pacemaker fitting.
+Wheels are available for Linux, Apple Silicon, and Intel Macs.
+
+## Add the GRACE fitting backend
+
+On Linux or Apple Silicon, install:
+
+```bash
+python -m pip install 'tensorpotential==0.6.2.1' 'numpy>=2,<2.4' 'tensorflow~=2.20.0' 'tf-keras~=2.20.0'
+```
+
+GRACE is published as `tensorpotential`. Use these versions together: this
+backend release requires NumPy below 2.4 for pacemaker fitting.
+TensorFlow 2.20 does not provide Intel Mac wheels.
 
 ## Check the installation
 
@@ -27,7 +37,7 @@ python -m pip check
 pacemaker --help
 ```
 
-GPU fitting requires a compatible NVIDIA driver. To run on the CPU, set
+On Linux, GPU fitting requires a compatible NVIDIA driver. To use the CPU, set
 `CUDA_VISIBLE_DEVICES=-1` before starting pacemaker.
 
 Continue with the [quick start](quickstart.md) to prepare data and run a fit.