From 7e98843d97c3b72af087d629b59bd1eb2902065e Mon Sep 17 00:00:00 2001
From: Sameer Kankute <sameer@berri.ai>
Date: Thu, 8 Jan 2026 10:57:07 +0530
Subject: [PATCH] Fix: Incomplete usage in response object passed

---
 litellm/llms/anthropic/chat/transformation.py | 15 ++---
 .../test_anthropic_chat_transformation.py     | 58 +++++++++++++++++++
 2 files changed, 66 insertions(+), 7 deletions(-)

diff --git a/litellm/llms/anthropic/chat/transformation.py b/litellm/llms/anthropic/chat/transformation.py
index c71edcdc2d1..57391c152cb 100644
--- a/litellm/llms/anthropic/chat/transformation.py
+++ b/litellm/llms/anthropic/chat/transformation.py
@@ -1265,14 +1265,15 @@ class AnthropicConfig(AnthropicModelInfo, BaseConfig):
             cache_creation_tokens=cache_creation_input_tokens,
             cache_creation_token_details=cache_creation_token_details,
         )
-        completion_token_details = (
-            CompletionTokensDetailsWrapper(
-                reasoning_tokens=token_counter(
-                    text=reasoning_content, count_response_tokens=True
-                )
-            )
+        # Always populate completion_token_details, not just when there's reasoning_content
+        reasoning_tokens = (
+            token_counter(text=reasoning_content, count_response_tokens=True)
             if reasoning_content
-            else None
+            else 0
+        )
+        completion_token_details = CompletionTokensDetailsWrapper(
+            reasoning_tokens=reasoning_tokens if reasoning_tokens > 0 else None,
+            text_tokens=completion_tokens - reasoning_tokens if reasoning_tokens > 0 else completion_tokens,
         )
         total_tokens = prompt_tokens + completion_tokens
 
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 9b6d1c6e178..5e58601d589 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
@@ -1754,3 +1754,61 @@ def test_transform_request_respects_user_max_tokens():
     )
 
     assert result["max_tokens"] == 1000
+
+
+def test_calculate_usage_completion_tokens_details_always_populated():
+    """
+    Test that completion_tokens_details is always populated in Usage object,
+    not just when there's reasoning_content.
+    
+    Fixes: https://github.com/BerriAI/litellm/issues/18772
+    Bug: completion_tokens_details was None for regular Claude responses without reasoning
+    """
+    config = AnthropicConfig()
+
+    # Test without reasoning_content - completion_tokens_details should still be populated
+    usage_object = {
+        "input_tokens": 37,
+        "output_tokens": 248,
+    }
+    usage = config.calculate_usage(usage_object=usage_object, reasoning_content=None)
+    
+    # completion_tokens_details should NOT be None
+    assert usage.completion_tokens_details is not None
+    assert usage.completion_tokens_details.reasoning_tokens is None
+    assert usage.completion_tokens_details.text_tokens == 248
+    assert usage.completion_tokens == 248
+    assert usage.prompt_tokens == 37
+    assert usage.total_tokens == 285
+
+
+def test_calculate_usage_completion_tokens_details_with_reasoning():
+    """
+    Test that completion_tokens_details correctly splits text_tokens and reasoning_tokens
+    when reasoning_content is present.
+    
+    Fixes: https://github.com/BerriAI/litellm/issues/18772
+    """
+    config = AnthropicConfig()
+
+    # Test with reasoning_content - should split tokens correctly
+    usage_object = {
+        "input_tokens": 100,
+        "output_tokens": 500,
+    }
+    # Simulating reasoning content that would count as ~50 tokens
+    reasoning_content = "Let me think about this step by step. " * 10  # Roughly 50 tokens
+    
+    usage = config.calculate_usage(
+        usage_object=usage_object, 
+        reasoning_content=reasoning_content
+    )
+    
+    # completion_tokens_details should be populated with both reasoning and text tokens
+    assert usage.completion_tokens_details is not None
+    assert usage.completion_tokens_details.reasoning_tokens is not None
+    assert usage.completion_tokens_details.reasoning_tokens > 0
+    # text_tokens should be total minus reasoning
+    expected_text_tokens = 500 - usage.completion_tokens_details.reasoning_tokens
+    assert usage.completion_tokens_details.text_tokens == expected_text_tokens
+    assert usage.completion_tokens == 500