diff --git a/src/magentic/chat.py b/src/magentic/chat.py
index 2e0dbab7..13e4e6aa 100644
--- a/src/magentic/chat.py
+++ b/src/magentic/chat.py
@@ -90,23 +90,37 @@ def add_assistant_message(self: Self, content: Any) -> Self:
         """Add an assistant message to the chat."""
         return self.add_message(AssistantMessage(content=content))
 
-    def submit(self: Self) -> Self:
+    def submit(self: Self, num_retries: int | None = None) -> Self:
         """Request an LLM message to be added to the chat."""
-        output_message: AssistantMessage[Any] = self.model.complete(
-            messages=self._messages,
-            functions=self._functions,
-            output_types=self._output_types,
-        )
-        return self.add_message(output_message)
+        retries = num_retries if num_retries is not None else 0
+        for attempt in range(retries + 1):
+            try:
+                output_message: AssistantMessage[Any] = self.model.complete(
+                    messages=self._messages,
+                    functions=self._functions,
+                    output_types=self._output_types,
+                )
+                return self.add_message(output_message)
+            except Exception as e:
+                if attempt == retries:
+                    raise
+                self._messages.append(UserMessage(content=f"Error: {str(e)}"))
 
-    async def asubmit(self: Self) -> Self:
+    async def asubmit(self: Self, num_retries: int | None = None) -> Self:
         """Async version of `submit`."""
-        output_message: AssistantMessage[Any] = await self.model.acomplete(
-            messages=self._messages,
-            functions=self._functions,
-            output_types=self._output_types,
-        )
-        return self.add_message(output_message)
+        retries = num_retries if num_retries is not None else 0
+        for attempt in range(retries + 1):
+            try:
+                output_message: AssistantMessage[Any] = await self.model.acomplete(
+                    messages=self._messages,
+                    functions=self._functions,
+                    output_types=self._output_types,
+                )
+                return self.add_message(output_message)
+            except Exception as e:
+                if attempt == retries:
+                    raise
+                self._messages.append(UserMessage(content=f"Error: {str(e)}"))
 
     def exec_function_call(self: Self) -> Self:
         """If the last message is a function call, execute it and add the result."""
diff --git a/src/magentic/prompt_function.py b/src/magentic/prompt_function.py
index 179c4ab8..efc12ea8 100644
--- a/src/magentic/prompt_function.py
+++ b/src/magentic/prompt_function.py
@@ -38,6 +38,7 @@ def __init__(
         functions: list[Callable[..., Any]] | None = None,
         stop: list[str] | None = None,
         model: ChatModel | None = None,
+        num_retries: int | None = None,
     ):
         self._signature = inspect.Signature(
             parameters=parameters,
@@ -49,6 +50,7 @@ def __init__(
         self._model = model
 
         self._return_types = list(split_union_type(return_type))
+        self._num_retries = num_retries
 
     @property
     def functions(self) -> list[Callable[..., Any]]:
@@ -78,13 +80,20 @@ class PromptFunction(BasePromptFunction[P, R], Generic[P, R]):
 
     def __call__(self, *args: P.args, **kwargs: P.kwargs) -> R:
         """Query the LLM with the formatted prompt template."""
-        message = self.model.complete(
-            messages=[UserMessage(content=self.format(*args, **kwargs))],
-            functions=self._functions,
-            output_types=self._return_types,
-            stop=self._stop,
-        )
-        return message.content
+        retries = self._num_retries if self._num_retries is not None else 0
+        for attempt in range(retries + 1):
+            try:
+                message = self.model.complete(
+                    messages=[UserMessage(content=self.format(*args, **kwargs))],
+                    functions=self._functions,
+                    output_types=self._return_types,
+                    stop=self._stop,
+                )
+                return message.content
+            except Exception as e:
+                if attempt == retries:
+                    raise
+                self._messages.append(UserMessage(content=f"Error: {str(e)}"))
 
 
 class AsyncPromptFunction(BasePromptFunction[P, R], Generic[P, R]):
@@ -92,13 +101,20 @@ class AsyncPromptFunction(BasePromptFunction[P, R], Generic[P, R]):
 
     async def __call__(self, *args: P.args, **kwargs: P.kwargs) -> R:
         """Asynchronously query the LLM with the formatted prompt template."""
-        message = await self.model.acomplete(
-            messages=[UserMessage(content=self.format(*args, **kwargs))],
-            functions=self._functions,
-            output_types=self._return_types,
-            stop=self._stop,
-        )
-        return message.content
+        retries = self._num_retries if self._num_retries is not None else 0
+        for attempt in range(retries + 1):
+            try:
+                message = await self.model.acomplete(
+                    messages=[UserMessage(content=self.format(*args, **kwargs))],
+                    functions=self._functions,
+                    output_types=self._return_types,
+                    stop=self._stop,
+                )
+                return message.content
+            except Exception as e:
+                if attempt == retries:
+                    raise
+                self._messages.append(UserMessage(content=f"Error: {str(e)}"))
 
 
 class PromptDecorator(Protocol):