diff --git a/python/packages/autogen-agentchat/src/autogen_agentchat/agents/_code_executor_agent.py b/python/packages/autogen-agentchat/src/autogen_agentchat/agents/_code_executor_agent.py
index ae9312cc7d67..8006164e5dec 100644
--- a/python/packages/autogen-agentchat/src/autogen_agentchat/agents/_code_executor_agent.py
+++ b/python/packages/autogen-agentchat/src/autogen_agentchat/agents/_code_executor_agent.py
@@ -164,6 +164,45 @@ async def run_code_executor_agent() -> None:
 
             asyncio.run(run_code_executor_agent())
 
+    In this example, we show how to set up a `CodeExecutorAgent` agent that uses the
+    :py:class:`~docker.types.DeviceRequest` to expose a GPU to the container for cuda-accelerated code execution.
+
+        .. code-block:: python
+
+            import asyncio
+            from autogen_agentchat.agents import CodeExecutorAgent
+            from autogen_agentchat.messages import TextMessage
+            from autogen_ext.code_executors.docker import DockerCommandLineCodeExecutor
+            from autogen_core import CancellationToken
+            from docker.types import DeviceRequest
+
+
+            async def run_code_executor_agent() -> None:
+                # Create a code executor agent that uses a Docker container to execute code.
+                code_executor = DockerCommandLineCodeExecutor(
+                    work_dir="coding", device_requests=[DeviceRequest(count=-1, capabilities=[["gpu"]])]
+                )
+                await code_executor.start()
+                code_executor_agent = CodeExecutorAgent("code_executor", code_executor=code_executor)
+
+                # Display the GPU information
+                task = TextMessage(
+                    content='''Here is some code
+            ```bash
+            nvidia-smi
+            ```
+            ''',
+                    source="user",
+                )
+                response = await code_executor_agent.on_messages([task], CancellationToken())
+                print(response.chat_message)
+
+                # Stop the code executor.
+                await code_executor.stop()
+
+
+            asyncio.run(run_code_executor_agent())
+
     In the following example, we show how to setup `CodeExecutorAgent` without `model_client` parameter for executing code blocks generated by other agents in a group chat using :py:class:`~autogen_ext.code_executors.docker.DockerCommandLineCodeExecutor`
 
         .. code-block:: python
diff --git a/python/packages/autogen-ext/src/autogen_ext/code_executors/docker/_docker_code_executor.py b/python/packages/autogen-ext/src/autogen_ext/code_executors/docker/_docker_code_executor.py
index fda23da0b349..aaef37c0b2b5 100644
--- a/python/packages/autogen-ext/src/autogen_ext/code_executors/docker/_docker_code_executor.py
+++ b/python/packages/autogen-ext/src/autogen_ext/code_executors/docker/_docker_code_executor.py
@@ -25,6 +25,8 @@
 from pydantic import BaseModel
 from typing_extensions import Self
 
+from docker.types import DeviceRequest
+
 from .._common import (
     CommandLineCodeResult,
     build_python_functions_file,
@@ -116,6 +118,7 @@ class DockerCommandLineCodeExecutor(CodeExecutor, Component[DockerCommandLineCod
         stop_container (bool, optional): If true, will automatically stop the
             container when stop is called, when the context manager exits or when
             the Python process exits with atext. Defaults to True.
+        device_requests (Optional[List[DeviceRequest]], optional): A list of device request instances to add to the container for exposing GPUs (e.g., [docker.types.DeviceRequest(count=-1, capabilities=[['gpu']])]). Defaults to None.
         functions (List[Union[FunctionWithRequirements[Any, A], Callable[..., Any]]]): A list of functions that are available to the code executor. Default is an empty list.
         functions_module (str, optional): The name of the module that will be created to store the functions. Defaults to "functions".
         extra_volumes (Optional[Dict[str, Dict[str, str]]], optional): A dictionary of extra volumes (beyond the work_dir) to mount to the container;
@@ -163,6 +166,7 @@ def __init__(
         bind_dir: Optional[Union[Path, str]] = None,
         auto_remove: bool = True,
         stop_container: bool = True,
+        device_requests: Optional[List[DeviceRequest]] = None,
         functions: Sequence[
             Union[
                 FunctionWithRequirements[Any, A],
@@ -229,6 +233,7 @@ def __init__(
         self._extra_hosts = extra_hosts if extra_hosts is not None else {}
         self._init_command = init_command
         self._delete_tmp_files = delete_tmp_files
+        self._device_requests = device_requests
 
         # Setup could take some time so we intentionally wait for the first code block to do it.
         if len(functions) > 0:
@@ -488,6 +493,7 @@ async def start(self) -> None:
             volumes={str(self.bind_dir.resolve()): {"bind": "/workspace", "mode": "rw"}, **self._extra_volumes},
             working_dir="/workspace",
             extra_hosts=self._extra_hosts,
+            device_requests=self._device_requests,
         )
         await asyncio.to_thread(self._container.start)