diff --git a/litellm/llms/snowflake/chat/transformation.py b/litellm/llms/snowflake/chat/transformation.py
index c64fc583edc..0a68805ba40 100644
--- a/litellm/llms/snowflake/chat/transformation.py
+++ b/litellm/llms/snowflake/chat/transformation.py
@@ -461,7 +461,12 @@ class SnowflakeConfig(SnowflakeBaseConfig, OpenAIGPTConfig):
                 "required": {"type": "any"},
                 "none": {"type": "none"},
             }
-            return mapping.get(tool_choice, {"type": "auto"})
+            if tool_choice not in mapping:
+                raise ValueError(
+                    "Unsupported tool_choice value for Snowflake Claude model: "
+                    f"{tool_choice!r}. Expected one of: auto, required, none."
+                )
+            return mapping[tool_choice]
         elif isinstance(tool_choice, dict):
             if tool_choice.get("type") == "function":
                 func: Final = tool_choice.get("function", {})
diff --git a/tests/test_litellm/llms/snowflake/chat/test_snowflake_chat_transformation.py b/tests/test_litellm/llms/snowflake/chat/test_snowflake_chat_transformation.py
index 25a961c3413..c37827d4ef8 100644
--- a/tests/test_litellm/llms/snowflake/chat/test_snowflake_chat_transformation.py
+++ b/tests/test_litellm/llms/snowflake/chat/test_snowflake_chat_transformation.py
@@ -117,6 +117,21 @@ class TestSnowflakeToolTransformation:
                 f"tool_choice='{value}' should pass through unchanged, got {transformed_request['tool_choice']}"
             )
 
+    def test_claude_rejects_invalid_string_tool_choice(self):
+        """
+        Test that invalid string tool_choice values are not silently downgraded to auto.
+        """
+        config = SnowflakeConfig()
+
+        with pytest.raises(ValueError, match="Unsupported tool_choice value"):
+            config.transform_request(
+                model="snowflake/claude-sonnet-4-5",
+                messages=[{"role": "user", "content": "Test"}],
+                optional_params={"tool_choice": "requried"},
+                litellm_params={},
+                headers={},
+            )
+
     def test_transform_response_with_tool_calls(self):
         """
         Test that standard OpenAI tool_calls response format is parsed correctly.