diff --git a/litellm/llms/bedrock.py b/litellm/llms/bedrock.py
index 894114559d..235c13c59c 100644
--- a/litellm/llms/bedrock.py
+++ b/litellm/llms/bedrock.py
@@ -952,7 +952,7 @@ def completion(
             original_response=json.dumps(response_body),
             additional_args={"complete_input_dict": data},
         )
-        print_verbose(f"raw model_response: {response}")
+        print_verbose(f"raw model_response: {response_body}")
         ## RESPONSE OBJECT
         outputText = "default"
         if provider == "ai21":
@@ -1065,6 +1065,7 @@ def completion(
             outputText = response_body.get("results")[0].get("outputText")
 
         response_metadata = response.get("ResponseMetadata", {})
+
         if response_metadata.get("HTTPStatusCode", 500) >= 400:
             raise BedrockError(
                 message=outputText,
@@ -1100,11 +1101,13 @@ def completion(
             prompt_tokens = response_metadata.get(
                 "x-amzn-bedrock-input-token-count", len(encoding.encode(prompt))
             )
+            _text_response = model_response["choices"][0]["message"].get("content", "")
             completion_tokens = response_metadata.get(
                 "x-amzn-bedrock-output-token-count",
                 len(
                     encoding.encode(
-                        model_response["choices"][0]["message"].get("content", "")
+                        _text_response,
+                        disallowed_special=(),
                     )
                 ),
             )
diff --git a/litellm/tests/test_custom_callback_input.py b/litellm/tests/test_custom_callback_input.py
index 0fd2cfabb5..2754ac6561 100644
--- a/litellm/tests/test_custom_callback_input.py
+++ b/litellm/tests/test_custom_callback_input.py
@@ -529,6 +529,7 @@ def test_chat_bedrock_stream():
 @pytest.mark.asyncio
 async def test_async_chat_bedrock_stream():
     try:
+        litellm.set_verbose = True
         customHandler = CompletionCustomHandler()
         litellm.callbacks = [customHandler]
         response = await litellm.acompletion(
diff --git a/litellm/tests/test_streaming.py b/litellm/tests/test_streaming.py
index d0d8a720a7..271a53dd49 100644
--- a/litellm/tests/test_streaming.py
+++ b/litellm/tests/test_streaming.py
@@ -518,7 +518,7 @@ async def test_acompletion_gemini_stream():
         litellm.set_verbose = True
         print("Streaming gemini response")
         messages = [
-            {"role": "system", "content": "You are a helpful assistant."},
+            # {"role": "system", "content": "You are a helpful assistant."},
             {
                 "role": "user",
                 "content": "What do you know?",
@@ -1297,6 +1297,7 @@ def test_completion_watsonx_stream():
     except Exception as e:
         pytest.fail(f"Error occurred: {e}")
 
+
 # test_completion_sagemaker_stream()
 
 
diff --git a/litellm/utils.py b/litellm/utils.py
index be14b13776..6b8aaaa5d8 100644
--- a/litellm/utils.py
+++ b/litellm/utils.py
@@ -410,10 +410,11 @@ class Message(OpenAIObject):
         except:
             # if using pydantic v1
             return self.dict()
-            
+
+
 class Delta(OpenAIObject):
-    tool_calls: Optional[List[ChatCompletionDeltaToolCall]]
-    
+    tool_calls: Optional[List[ChatCompletionDeltaToolCall]] = None
+
     def __init__(
         self,
         content=None,
@@ -475,7 +476,7 @@ class Choices(OpenAIObject):
         )  # set finish_reason for all responses
         self.index = index
         if message is None:
-            self.message = Message(content=None)
+            self.message = Message()
         else:
             if isinstance(message, Message):
                 self.message = message