diff --git a/litellm/main.py b/litellm/main.py
index 5a5c1adcc98..cc020d523b3 100644
--- a/litellm/main.py
+++ b/litellm/main.py
@@ -5480,11 +5480,6 @@ def completion(
     ### VALIDATE Request ###
     if model is None:
         raise ValueError("model param not passed in.")
-    supplemental_params: Final = _supplemental_provider_params(
-        _REQUEST_KWARGS_ADAPTER.validate_python({**kwargs, "extra_headers": extra_headers}), model
-    )
-    for dropped_key in supplemental_params.dropped_keys:
-        kwargs.pop(dropped_key)
     # validate messages
     messages = validate_and_fix_openai_messages(messages=messages)
     tools = validate_and_fix_openai_tools(tools=tools)
@@ -5496,6 +5491,11 @@ def completion(
 
     ######### unpacking kwargs #####################
     args: Final = _locals_snapshot(locals())
+    supplemental_params: Final = _supplemental_provider_params(
+        _REQUEST_KWARGS_ADAPTER.validate_python({**kwargs, "extra_headers": extra_headers}), model
+    )
+    for dropped_key in supplemental_params.dropped_keys:
+        kwargs.pop(dropped_key)
 
     # Set by the responses->completion fallback so completion() does not bridge
     # back to the Responses API: that round-trip mutually recurses forever for a
diff --git a/tests/e2e/llm_translation/test_gemini_tool_call_id_length_e2e.py b/tests/e2e/llm_translation/test_gemini_tool_call_id_length_e2e.py
index 883b6c08aa2..64fc3aa72af 100644
--- a/tests/e2e/llm_translation/test_gemini_tool_call_id_length_e2e.py
+++ b/tests/e2e/llm_translation/test_gemini_tool_call_id_length_e2e.py
@@ -25,7 +25,7 @@ from passthrough_client import PassthroughClient
 
 pytestmark = pytest.mark.e2e
 
-BACKEND_MODEL: Final = "gemini/gemini-3.5-flash"
+BACKEND_MODEL: Final = "gemini/gemini-3.8-flash"
 GEMINI_API_KEY: Final = "os.environ/GEMINI_API_KEY"
 # bedrock-agentcore HarnessToolUseId shape in botocore 1.43.6, read 2026-10-20
 AGENTCORE_TOOL_USE_ID: Final = re.compile(r"[a-zA-Z0-9_-]{1,64}")
@@ -117,7 +117,7 @@ class TestGeminiToolCallIdLength:
             messages=[_user_message()],
             tools=[WEATHER_TOOL],
             tool_choice="required",
-            max_tokens=128,
+            max_tokens=2048,
             extra_body=NO_PROXY_CACHE,
         )
 
@@ -138,7 +138,7 @@ class TestGeminiToolCallIdLength:
                 messages=[_user_message()],
                 tools=[WEATHER_TOOL],
                 tool_choice="required",
-                max_tokens=128,
+                max_tokens=2048,
                 stream=True,
                 extra_body=NO_PROXY_CACHE,
             )
@@ -164,7 +164,7 @@ class TestGeminiToolCallIdLength:
             messages=[question],
             tools=[WEATHER_TOOL],
             tool_choice="required",
-            max_tokens=128,
+            max_tokens=2048,
             extra_body=NO_PROXY_CACHE,
         )
         calls: Final = _tool_calls(first)
@@ -178,7 +178,7 @@ class TestGeminiToolCallIdLength:
             messages=[question, assistant, *(_tool_result(call) for call in calls)],
             tools=[WEATHER_TOOL],
             tool_choice="none",
-            max_tokens=128,
+            max_tokens=2048,
             extra_body=NO_PROXY_CACHE,
         )