diff --git a/litellm/litellm_core_utils/exception_mapping_utils.py b/litellm/litellm_core_utils/exception_mapping_utils.py
index ee179b582cf..4a25eb218c0 100644
--- a/litellm/litellm_core_utils/exception_mapping_utils.py
+++ b/litellm/litellm_core_utils/exception_mapping_utils.py
@@ -811,6 +811,24 @@ def _map_openai_like_exception(
             )
 
 
+_BEDROCK_MANTLE_CONTEXT_WINDOW_PATTERN: Final = re.compile(r"prompt tokens \((\d+)\) exceed model maximum \((\d+)\)")
+
+
+def _get_bedrock_mantle_context_window_message(error_str: str) -> str | None:
+    """
+    Mantle reports context overflow as a structured validation error rather than
+    the plain-text patterns Bedrock itself uses, so it needs its own detection and a
+    message clients recognize as context overflow (litellm/litellm#36546).
+    """
+    if "invalid_request_error" not in error_str and "validation_error" not in error_str:
+        return None
+    match = _BEDROCK_MANTLE_CONTEXT_WINDOW_PATTERN.search(error_str)
+    if match is None:
+        return None
+    prompt_tokens, max_tokens = match.groups()
+    return f"prompt is too long: {prompt_tokens} tokens > {max_tokens} maximum"
+
+
 def _map_bedrock_exception(
     *,
     model: str,
@@ -821,6 +839,14 @@ def _map_bedrock_exception(
     exception_provider: str,
     extra_information: str,
 ) -> None:
+    if custom_llm_provider == "bedrock_mantle":
+        mantle_context_window_message = _get_bedrock_mantle_context_window_message(error_str)
+        if mantle_context_window_message is not None:
+            raise ContextWindowExceededError(
+                message=mantle_context_window_message,
+                model=model,
+                llm_provider=custom_llm_provider,
+            )
     if (
         "too many tokens" in error_str
         or "expected maxLength:" in error_str
diff --git a/tests/test_litellm/litellm_core_utils/test_exception_mapping_utils.py b/tests/test_litellm/litellm_core_utils/test_exception_mapping_utils.py
index 904675d4adc..38f46b26eea 100644
--- a/tests/test_litellm/litellm_core_utils/test_exception_mapping_utils.py
+++ b/tests/test_litellm/litellm_core_utils/test_exception_mapping_utils.py
@@ -785,3 +785,27 @@ def test_bedrock_mantle_400_maps_to_bad_request():
 
     assert excinfo.value.status_code == 400
     assert "Invalid 'input'" in excinfo.value.message
+    assert type(excinfo.value) is litellm.BadRequestError
+
+
+def test_bedrock_mantle_context_overflow_maps_to_context_window_exceeded():
+    from litellm.llms.base_llm.chat.transformation import BaseLLMException
+
+    original_exception = BaseLLMException(
+        status_code=400,
+        message=(
+            '{"error":{"code":"validation_error",'
+            '"message":"prompt tokens (1055489) exceed model maximum (1050000) for openai.gpt-5.6-sol",'
+            '"param":null,"type":"invalid_request_error"}}'
+        ),
+    )
+
+    with pytest.raises(litellm.ContextWindowExceededError) as excinfo:
+        exception_type(
+            model="openai.gpt-5.6-sol",
+            original_exception=original_exception,
+            custom_llm_provider="bedrock_mantle",
+        )
+
+    assert excinfo.value.status_code == 400
+    assert "prompt is too long: 1055489 tokens > 1050000 maximum" in excinfo.value.message