diff --git a/litellm/llms/anthropic/chat/transformation.py b/litellm/llms/anthropic/chat/transformation.py
index 907a6d85a1c..34d259d2586 100644
--- a/litellm/llms/anthropic/chat/transformation.py
+++ b/litellm/llms/anthropic/chat/transformation.py
@@ -5,6 +5,7 @@ from collections.abc import Mapping
 from typing import TYPE_CHECKING, Any, Final, NoReturn, cast
 
 import httpx
+from pydantic import ValidationError
 
 import litellm
 from litellm.constants import (
@@ -39,6 +40,7 @@ from litellm.types.llms.anthropic import (
     AnthropicMessagesTool,
     AnthropicMessagesToolChoice,
     AnthropicOutputSchema,
+    AnthropicOutputTokensDetails,
     AnthropicSystemMessageContent,
     AnthropicThinkingParam,
     AnthropicWebSearchTool,
@@ -2129,10 +2131,11 @@ class AnthropicConfig(AnthropicModelInfo, BaseConfig):
         details: Final = usage_object.get("output_tokens_details")
         if not isinstance(details, Mapping):
             return None
-        thinking_tokens: Final = cast(Mapping[str, object], details).get("thinking_tokens")
-        if isinstance(thinking_tokens, bool) or not isinstance(thinking_tokens, (int, float)):
+        try:
+            parsed_details: Final = AnthropicOutputTokensDetails.model_validate(details)
+        except ValidationError:
             return None
-        return int(thinking_tokens)
+        return parsed_details.thinking_tokens
 
     def calculate_usage(
         self,
diff --git a/litellm/types/llms/anthropic.py b/litellm/types/llms/anthropic.py
index bb861030d86..ee0d2c098f3 100644
--- a/litellm/types/llms/anthropic.py
+++ b/litellm/types/llms/anthropic.py
@@ -609,6 +609,12 @@ class AnthropicResponseUsageBlock(BaseModel):
     output_tokens: int
 
 
+class AnthropicOutputTokensDetails(BaseModel):
+    model_config = ConfigDict(extra="allow", strict=True)
+
+    thinking_tokens: int | None = None
+
+
 AnthropicFinishReason = Literal["end_turn", "max_tokens", "stop_sequence", "tool_use"]
 
 
diff --git a/tests/test_litellm/llms/anthropic/chat/test_anthropic_chat_transformation.py b/tests/test_litellm/llms/anthropic/chat/test_anthropic_chat_transformation.py
index 6fc61da326a..2c59ee59df8 100644
--- a/tests/test_litellm/llms/anthropic/chat/test_anthropic_chat_transformation.py
+++ b/tests/test_litellm/llms/anthropic/chat/test_anthropic_chat_transformation.py
@@ -2477,7 +2477,7 @@ def test_calculate_usage_reported_thinking_tokens_override_text_estimate():
 
 @pytest.mark.parametrize(
     "output_tokens_details",
-    [None, {}, {"thinking_tokens": None}],
+    [None, {}, {"thinking_tokens": None}, {"thinking_tokens": "180"}, {"thinking_tokens": True}],
 )
 def test_calculate_usage_falls_back_to_estimate_without_reported_thinking_tokens(output_tokens_details):
     """Older responses have no thinking_tokens, so the visible-text estimate must still apply."""