diff --git a/litellm/llms/vertex_httpx.py b/litellm/llms/vertex_httpx.py
index 31910b7ea39..d56237b73d1 100644
--- a/litellm/llms/vertex_httpx.py
+++ b/litellm/llms/vertex_httpx.py
@@ -603,15 +603,15 @@ class VertexLLM(BaseLLM):
 
                 ## GET USAGE ##
                 usage = litellm.Usage(
-                    prompt_tokens=completion_response["usageMetadata"][
-                        "promptTokenCount"
-                    ],
+                    prompt_tokens=completion_response["usageMetadata"].get(
+                        "promptTokenCount", 0
+                    ),
                     completion_tokens=completion_response["usageMetadata"].get(
                         "candidatesTokenCount", 0
                     ),
-                    total_tokens=completion_response["usageMetadata"][
-                        "totalTokenCount"
-                    ],
+                    total_tokens=completion_response["usageMetadata"].get(
+                        "totalTokenCount", 0
+                    ),
                 )
 
                 setattr(model_response, "usage", usage)
@@ -647,15 +647,15 @@ class VertexLLM(BaseLLM):
 
                 ## GET USAGE ##
                 usage = litellm.Usage(
-                    prompt_tokens=completion_response["usageMetadata"][
-                        "promptTokenCount"
-                    ],
+                    prompt_tokens=completion_response["usageMetadata"].get(
+                        "promptTokenCount", 0
+                    ),
                     completion_tokens=completion_response["usageMetadata"].get(
                         "candidatesTokenCount", 0
                     ),
-                    total_tokens=completion_response["usageMetadata"][
-                        "totalTokenCount"
-                    ],
+                    total_tokens=completion_response["usageMetadata"].get(
+                        "totalTokenCount", 0
+                    ),
                 )
 
                 setattr(model_response, "usage", usage)
@@ -705,11 +705,15 @@ class VertexLLM(BaseLLM):
 
             ## GET USAGE ##
             usage = litellm.Usage(
-                prompt_tokens=completion_response["usageMetadata"]["promptTokenCount"],
+                prompt_tokens=completion_response["usageMetadata"].get(
+                    "promptTokenCount", 0
+                ),
                 completion_tokens=completion_response["usageMetadata"].get(
                     "candidatesTokenCount", 0
                 ),
-                total_tokens=completion_response["usageMetadata"]["totalTokenCount"],
+                total_tokens=completion_response["usageMetadata"].get(
+                    "totalTokenCount", 0
+                ),
             )
 
             setattr(model_response, "usage", usage)
@@ -1340,11 +1344,15 @@ class ModelResponseIterator:
 
             if "usageMetadata" in processed_chunk:
                 usage = ChatCompletionUsageBlock(
-                    prompt_tokens=processed_chunk["usageMetadata"]["promptTokenCount"],
+                    prompt_tokens=processed_chunk["usageMetadata"].get(
+                        "promptTokenCount", 0
+                    ),
                     completion_tokens=processed_chunk["usageMetadata"].get(
                         "candidatesTokenCount", 0
                     ),
-                    total_tokens=processed_chunk["usageMetadata"]["totalTokenCount"],
+                    total_tokens=processed_chunk["usageMetadata"].get(
+                        "totalTokenCount", 0
+                    ),
                 )
 
             returned_chunk = GenericStreamingChunk(
diff --git a/litellm/types/llms/vertex_ai.py b/litellm/types/llms/vertex_ai.py
index 17fc26d60ef..f07767a2dd4 100644
--- a/litellm/types/llms/vertex_ai.py
+++ b/litellm/types/llms/vertex_ai.py
@@ -239,8 +239,8 @@ class PromptFeedback(TypedDict):
 
 
 class UsageMetadata(TypedDict, total=False):
-    promptTokenCount: Required[int]
-    totalTokenCount: Required[int]
+    promptTokenCount: int
+    totalTokenCount: int
     candidatesTokenCount: int