diff --git a/litellm/llms/vertex_ai/context_caching/vertex_ai_context_caching.py b/litellm/llms/vertex_ai/context_caching/vertex_ai_context_caching.py
index 049336905ef..f507d2dfb97 100644
--- a/litellm/llms/vertex_ai/context_caching/vertex_ai_context_caching.py
+++ b/litellm/llms/vertex_ai/context_caching/vertex_ai_context_caching.py
@@ -1,6 +1,7 @@
 from typing import Final, Literal
 
 import httpx
+from pydantic import TypeAdapter
 
 import litellm
 from litellm._logging import verbose_logger
@@ -14,6 +15,7 @@ from litellm.llms.custom_httpx.http_handler import (
 )
 from litellm.llms.openai.openai import AllMessageValues
 from litellm.types.llms.vertex_ai import (
+    VERTEX_AI_CACHED_CONTENT_KEY,
     CachedContentListAllResponseBody,
     VertexAICachedContentCreation,
     VertexAICachedContentResponseObject,
@@ -31,7 +33,6 @@ from .transformation import (
 local_cache_obj: Final = Cache(type=LiteLLMCacheType.LOCAL)  # only used for calling 'get_cache_key' function
 
 MAX_PAGINATION_PAGES: Final = 100  # Reasonable upper bound for pagination
-VERTEX_AI_CACHED_CONTENT_KEY: Final = "vertex_ai_cached_content"
 
 
 class ContextCachingEndpoints(VertexBase):
@@ -426,7 +427,9 @@ class ContextCachingEndpoints(VertexBase):
             raise VertexAIError(status_code=408, message="Timeout error occurred.")
 
         raw_response_cached: Final = response.json()
-        cached_content_response_obj: Final = VertexAICachedContentResponseObject(**raw_response_cached)
+        cached_content_response_obj: Final = TypeAdapter(VertexAICachedContentResponseObject).validate_python(
+            raw_response_cached
+        )
         usage_metadata: Final = cached_content_response_obj.get("usageMetadata", {})
         cached_content_creation: Final = VertexAICachedContentCreation(
             name=cached_content_response_obj["name"],
@@ -588,7 +591,9 @@ class ContextCachingEndpoints(VertexBase):
             raise VertexAIError(status_code=408, message="Timeout error occurred.")
 
         raw_response_cached: Final = response.json()
-        cached_content_response_obj: Final = VertexAICachedContentResponseObject(**raw_response_cached)
+        cached_content_response_obj: Final = TypeAdapter(VertexAICachedContentResponseObject).validate_python(
+            raw_response_cached
+        )
         usage_metadata: Final = cached_content_response_obj.get("usageMetadata", {})
         cached_content_creation: Final = VertexAICachedContentCreation(
             name=cached_content_response_obj["name"],
diff --git a/litellm/llms/vertex_ai/gemini/vertex_and_google_ai_studio_gemini.py b/litellm/llms/vertex_ai/gemini/vertex_and_google_ai_studio_gemini.py
index 51de4fa3f69..ddf1c9af0d4 100644
--- a/litellm/llms/vertex_ai/gemini/vertex_and_google_ai_studio_gemini.py
+++ b/litellm/llms/vertex_ai/gemini/vertex_and_google_ai_studio_gemini.py
@@ -51,6 +51,7 @@ from litellm.types.llms.openai import (
     OpenAIChatCompletionFinishReason,
 )
 from litellm.types.llms.vertex_ai import (
+    VERTEX_AI_CACHED_CONTENT_KEY,
     VERTEX_AI_PROVIDER_METADATA_FIELDS,
     VERTEX_CREDENTIALS_TYPES,
     Candidates,
@@ -142,15 +143,17 @@ def _add_cache_creation_usage(usage: Usage, creation: VertexAICachedContentCreat
         if usage.prompt_tokens_details is not None
         else PromptTokensDetailsWrapper()
     )
-    cache_read_tokens: Final = getattr(usage, "_cache_read_input_tokens", 0) or 0
+    cache_read_tokens: Final = prompt_tokens_details.cached_tokens or 0
+    server_tool_use: Final = usage.server_tool_use if hasattr(usage, "server_tool_use") else None
+    cost: Final = usage.cost if hasattr(usage, "cost") else None
     return Usage(
         prompt_tokens=usage.prompt_tokens + creation_tokens,
         completion_tokens=usage.completion_tokens,
         total_tokens=usage.total_tokens + creation_tokens,
         prompt_tokens_details=prompt_tokens_details,
         completion_tokens_details=usage.completion_tokens_details,
-        server_tool_use=getattr(usage, "server_tool_use", None),
-        cost=getattr(usage, "cost", None),
+        server_tool_use=server_tool_use,
+        cost=cost,
         cache_creation_input_tokens=creation_tokens,
         **({"cache_read_input_tokens": cache_read_tokens} if cache_read_tokens > 0 else {}),
     )
@@ -2505,8 +2508,6 @@ class VertexGeminiConfig(VertexAIBaseConfig, BaseConfig):
                 ) = VertexGeminiConfig._process_candidates(_candidates, model_response, logging_obj.optional_params)
 
             base_usage: Final = VertexGeminiConfig._calculate_usage(completion_response=completion_response)
-            from ..context_caching.vertex_ai_context_caching import VERTEX_AI_CACHED_CONTENT_KEY
-
             cached_content_creation: Final = logging_obj.model_call_details.get(VERTEX_AI_CACHED_CONTENT_KEY)
             usage: Final = (
                 _add_cache_creation_usage(base_usage, cast(VertexAICachedContentCreation, cached_content_creation))
@@ -3269,8 +3270,6 @@ class ModelResponseIterator:
         base_usage: Final = VertexGeminiConfig._calculate_usage(
             completion_response=processed_chunk,
         )
-        from ..context_caching.vertex_ai_context_caching import VERTEX_AI_CACHED_CONTENT_KEY
-
         cached_content_creation: Final = self.logging_obj.model_call_details.get(VERTEX_AI_CACHED_CONTENT_KEY)
         usage: Final = (
             _add_cache_creation_usage(base_usage, cast(VertexAICachedContentCreation, cached_content_creation))
diff --git a/litellm/types/llms/vertex_ai.py b/litellm/types/llms/vertex_ai.py
index 5ff792d6310..ca74741af16 100644
--- a/litellm/types/llms/vertex_ai.py
+++ b/litellm/types/llms/vertex_ai.py
@@ -576,6 +576,9 @@ class VertexAICachedContentCreation(TypedDict):
     expire_time: ReadOnly[str | None]
 
 
+VERTEX_AI_CACHED_CONTENT_KEY: Final = "vertex_ai_cached_content"
+
+
 class TaskTypeEnum(Enum):
     TASK_TYPE_UNSPECIFIED = "TASK_TYPE_UNSPECIFIED"
     RETRIEVAL_QUERY = "RETRIEVAL_QUERY"