diff --git a/docs/ramalama-run.1.md b/docs/ramalama-run.1.md
index 09d1c56a..d845298a 100644
--- a/docs/ramalama-run.1.md
+++ b/docs/ramalama-run.1.md
@@ -31,6 +31,15 @@ path of the authentication file for OCI registries
 #### **--ctx-size**, **-c**
 size of the prompt context (default: 2048, 0 = loaded from model)
 
+#### **--device**
+Add a host device to the container. Optional permissions parameter  can
+be  used  to  specify device permissions by combining r for read, w for
+write, and m for mknod(2).
+
+Example: --device=/dev/dri/renderD128:/dev/xvdc:rwm
+
+The device specifiaction is passed directly to the underlying container engine.  See documentation of the supported container engine for more information.
+
 #### **--help**, **-h**
 show this help message and exit
 
@@ -43,6 +52,24 @@ name of the container to run the Model in
 #### **--network**=*none*
 set the network mode for the container
 
+#### **--privileged**
+By  default, RamaLama containers are unprivileged (=false) and cannot, for
+example, modify parts of the operating system. This is  because  by  de‐
+fault  a  container is only allowed limited access to devices. A "privi‐
+leged" container is given the same access to devices as the user launch‐
+ing the container, with the exception of virtual consoles  (/dev/tty\d+)
+when running in systemd mode (--systemd=always).
+
+A  privileged container turns off the security features that isolate the
+container from the host. Dropped Capabilities,  limited  devices,  read-
+only  mount points, Apparmor/SELinux separation, and Seccomp filters are
+all disabled.  Due to the disabled  security  features,  the  privileged
+field  should  almost never be set as containers can easily break out of
+confinement.
+
+Containers running in a user namespace (e.g., rootless containers)  can‐
+not have more privileges than the user that launched them.
+
 #### **--seed**=
 Specify seed rather than using random seed model interaction
 
diff --git a/docs/ramalama-serve.1.md b/docs/ramalama-serve.1.md
index 837fdad5..a33074a6 100644
--- a/docs/ramalama-serve.1.md
+++ b/docs/ramalama-serve.1.md
@@ -46,6 +46,15 @@ The default is TRUE. The --nocontainer option forces this option to False.
 
 Use the `ramalama stop` command to stop the container running the served ramalama Model.
 
+#### **--device**
+Add a host device to the container. Optional permissions parameter  can
+be  used  to  specify device permissions by combining r for read, w for
+write, and m for mknod(2).
+
+Example: --device=/dev/dri/renderD128:/dev/xvdc:rwm
+
+The device specifiaction is passed directly to the underlying container engine.  See documentation of the supported container engine for more information.
+
 #### **--generate**=type
 Generate specified configuration format for running the AI Model as a service
 
@@ -70,6 +79,24 @@ set the network mode for the container
 #### **--port**, **-p**
 port for AI Model server to listen on
 
+#### **--privileged**
+By  default, RamaLama containers are unprivileged (=false) and cannot, for
+example, modify parts of the operating system. This is  because  by  de‐
+fault  a  container is only allowed limited access to devices. A "privi‐
+leged" container is given the same access to devices as the user launch‐
+ing the container, with the exception of virtual consoles  (/dev/tty\d+)
+when running in systemd mode (--systemd=always).
+
+A  privileged container turns off the security features that isolate the
+container from the host. Dropped Capabilities,  limited  devices,  read-
+only  mount points, Apparmor/SELinux separation, and Seccomp filters are
+all disabled.  Due to the disabled  security  features,  the  privileged
+field  should  almost never be set as containers can easily break out of
+confinement.
+
+Containers running in a user namespace (e.g., rootless containers)  can‐
+not have more privileges than the user that launched them.
+
 #### **--seed**=
 Specify seed rather than using random seed model interaction
 
diff --git a/ramalama/cli.py b/ramalama/cli.py
index 58677046..eed72ceb 100644
--- a/ramalama/cli.py
+++ b/ramalama/cli.py
@@ -801,6 +801,11 @@ def _run(parser):
         default=config.get('ctx_size', 2048),
         help="size of the prompt context (0 = loaded from model)",
     )
+    parser.add_argument("--device",
+        dest="device",
+        action='append',
+        type=str,
+        help="Device to leak in to the running container")
     parser.add_argument("-n", "--name", dest="name", help="name of container in which the Model will be run")
     # Disable network access by default, and give the option to pass any supported network mode into
     # podman if needed:
@@ -811,7 +816,12 @@ def _run(parser):
         default="none",
         help="set the network mode for the container",
     )
-
+    parser.add_argument(
+        "--privileged",
+        dest="privileged",
+        action="store_true",
+        help="give extended privileges to container"
+    )
     parser.add_argument("--seed", help="override random seed")
     parser.add_argument(
         "--temp", default=config.get('temp', "0.8"), help="temperature of the response from the AI model"
diff --git a/ramalama/model.py b/ramalama/model.py
index 768bf7d5..c4ba9c86 100644
--- a/ramalama/model.py
+++ b/ramalama/model.py
@@ -200,13 +200,10 @@ def setup_container(self, args):
         if hasattr(args, "port"):
             conman_args += ["-p", f"{args.port}:{args.port}"]
 
-        # Check for env var RAMALAMA_DEVICE to explicitly declare the GPU device path
-        device_override = 0
-        gpu_device = os.environ.get("RAMALAMA_DEVICE")
-        if gpu_device:
-            conman_args += ["--device", gpu_device]
-            device_override = 1
-        if device_override != 1:
+        if args.device:
+            for device_arg in args.device:
+                conman_args += ["--device", device_arg]
+        else:
             if (sys.platform == "darwin" and os.path.basename(args.engine) != "docker") or os.path.exists("/dev/dri"):
                 conman_args += ["--device", "/dev/dri"]