diff --git a/litellm/llms/xai/chat/transformation.py b/litellm/llms/xai/chat/transformation.py
index 91bf697487d..33ee727dfab 100644
--- a/litellm/llms/xai/chat/transformation.py
+++ b/litellm/llms/xai/chat/transformation.py
@@ -118,6 +118,7 @@ class XAIChatConfig(OpenAIGPTConfig):
         base_openai_params: Final = [
             "logit_bias",
             "logprobs",
+            "max_completion_tokens",
             "max_tokens",
             "n",
             "parallel_tool_calls",
diff --git a/tests/test_litellm/llms/xai/test_xai_chat_transformation.py b/tests/test_litellm/llms/xai/test_xai_chat_transformation.py
index 290cd3dcb3a..3fd666e4f50 100644
--- a/tests/test_litellm/llms/xai/test_xai_chat_transformation.py
+++ b/tests/test_litellm/llms/xai/test_xai_chat_transformation.py
@@ -90,6 +90,16 @@ class TestXAIReasoningTokenFolding:
         assert response.usage.total_tokens == 999
 
 
+def test_max_completion_tokens_is_accepted_and_mapped_to_max_tokens() -> None:
+    optional_params = litellm.get_optional_params(
+        model="grok-4.20",
+        custom_llm_provider="xai",
+        max_completion_tokens=64,
+    )
+    assert optional_params["max_tokens"] == 64, optional_params
+    assert "max_completion_tokens" not in optional_params, optional_params
+
+
 class TestXAIParallelToolCalls:
     """Test suite for XAI parallel tool calls functionality."""