From 7bf38c6312b19551331c3b698c1b6631f552db78 Mon Sep 17 00:00:00 2001
From: Gennaro Malafronte <8798987+malafronte@users.noreply.github.com>
Date: Thu, 23 Apr 2026 00:16:00 +0200
Subject: [PATCH] fix(anthropic): strip leaked think tags from messages

---
 litellm/llms/anthropic/chat/transformation.py |  3 ++
 .../adapters/transformation.py                |  8 +++--
 .../messages/handler.py                       | 34 ++++++++++---------
 3 files changed, 27 insertions(+), 18 deletions(-)

diff --git a/litellm/llms/anthropic/chat/transformation.py b/litellm/llms/anthropic/chat/transformation.py
index 31b9bad3395..8263ea4cf01 100644
--- a/litellm/llms/anthropic/chat/transformation.py
+++ b/litellm/llms/anthropic/chat/transformation.py
@@ -1667,6 +1667,9 @@ class AnthropicConfig(AnthropicModelInfo, BaseConfig):
                 if thinking_content is not None:
                     reasoning_content += thinking_content
 
+        if "</think>" in text_content:
+            text_content = text_content.split("</think>", 1)[1].lstrip()
+
         return (
             text_content,
             citations,
diff --git a/litellm/llms/anthropic/experimental_pass_through/adapters/transformation.py b/litellm/llms/anthropic/experimental_pass_through/adapters/transformation.py
index 23aed56e65f..e3c62dc9ea7 100644
--- a/litellm/llms/anthropic/experimental_pass_through/adapters/transformation.py
+++ b/litellm/llms/anthropic/experimental_pass_through/adapters/transformation.py
@@ -1266,10 +1266,14 @@ class LiteLLMAnthropicMessagesAdapter:
                 )
 
             # Handle text content
-            if choice.message.content is not None:
+            text_content = choice.message.content
+            if isinstance(text_content, str) and "</think>" in text_content:
+                text_content = text_content.split("</think>", 1)[1].lstrip()
+
+            if text_content:
                 new_content.append(
                     AnthropicResponseContentBlockText(
-                        type="text", text=choice.message.content
+                        type="text", text=text_content
                     ).model_dump()
                 )
             # Handle tool calls (in parallel to text content)
diff --git a/litellm/llms/anthropic/experimental_pass_through/messages/handler.py b/litellm/llms/anthropic/experimental_pass_through/messages/handler.py
index 0c59e812e0b..583f0e8b01b 100644
--- a/litellm/llms/anthropic/experimental_pass_through/messages/handler.py
+++ b/litellm/llms/anthropic/experimental_pass_through/messages/handler.py
@@ -454,20 +454,22 @@ def anthropic_messages_handler(
                 "display": "summarized",
             }
 
-    return base_llm_http_handler.anthropic_messages_handler(
-        model=model,
-        messages=messages,
-        anthropic_messages_provider_config=anthropic_messages_provider_config,
-        anthropic_messages_optional_request_params=dict(
-            anthropic_messages_optional_request_params
-        ),
-        _is_async=is_async,
-        client=client,
-        custom_llm_provider=custom_llm_provider,
-        litellm_params=litellm_params,
-        logging_obj=litellm_logging_obj,
-        api_key=api_key,
-        api_base=api_base,
-        stream=stream,
-        kwargs=kwargs,
+    return _sanitize_think_tag_text_blocks(
+        base_llm_http_handler.anthropic_messages_handler(
+            model=model,
+            messages=messages,
+            anthropic_messages_provider_config=anthropic_messages_provider_config,
+            anthropic_messages_optional_request_params=dict(
+                anthropic_messages_optional_request_params
+            ),
+            _is_async=is_async,
+            client=client,
+            custom_llm_provider=custom_llm_provider,
+            litellm_params=litellm_params,
+            logging_obj=litellm_logging_obj,
+            api_key=api_key,
+            api_base=api_base,
+            stream=stream,
+            kwargs=kwargs,
+        )
     )