diff --git a/libs/partners/anthropic/langchain_anthropic/chat_models.py b/libs/partners/anthropic/langchain_anthropic/chat_models.py
index ca0addb5f6..f9541d7052 100644
--- a/libs/partners/anthropic/langchain_anthropic/chat_models.py
+++ b/libs/partners/anthropic/langchain_anthropic/chat_models.py
@@ -993,6 +993,11 @@ def _supports_mid_conversation_system_messages(model: object) -> bool:
     )
 
 
+def _supports_forced_tool_choice(model: str) -> bool:
+    """Return whether the model accepts `tool_choice` types `any` and `tool`."""
+    return not model.startswith(("claude-fable-5-1", "claude-opus-5-5"))
+
+
 def _is_direct_anthropic_llm_type(llm_type: object) -> bool:
     """Return whether an `_llm_type` reaches Claude via the direct Anthropic API.
 
@@ -2394,9 +2399,11 @@ class ChatAnthropic(BaseChatModel):
     ) -> Runnable[LanguageModelInput, BaseMessage]:
         thinking_admonition = (
             "You are attempting to use structured output via forced tool calling, "
-            "which is not guaranteed when `thinking` is enabled. This method will "
-            "raise an OutputParserException if tool calls are not generated. Consider "
-            "disabling `thinking` or adjust your prompt to ensure the tool is called."
+            "which is not supported when `thinking` is enabled or on "
+            f"{self.model}. This method will raise an OutputParserException if tool "
+            "calls are not generated. Consider `method='json_schema'`, disabling "
+            "`thinking` where supported, or adjusting your prompt to ensure the "
+            "tool is called."
         )
         warnings.warn(thinking_admonition, stacklevel=2)
         llm = self.bind_tools(
@@ -2667,7 +2674,10 @@ class ChatAnthropic(BaseChatModel):
             method: The structured output method to use. Options are:
 
                 - `'function_calling'` (default): Use forced tool calling to get
-                    structured output.
+                    structured output. When `thinking` is enabled, or on models
+                    that don't support forced tool use (Claude Opus 5.5, Claude
+                    Fable 5.1), the tool call isn't forced, and a missing tool
+                    call raises `OutputParserException`.
                 - `'json_schema'`: Use Claude's dedicated
                     [structured output](https://platform.claude.com/docs/en/build-with-claude/structured-outputs)
                     feature.
@@ -2717,15 +2727,16 @@ class ChatAnthropic(BaseChatModel):
             warnings.warn(warning_message, stacklevel=2)
             method = "json_schema"
 
+        # TODO: make `method="json_schema"` the default in a future release.
         if method == "function_calling":
             formatted_tool = cast(AnthropicTool, convert_to_anthropic_tool(schema))
             # The result of convert_to_anthropic_tool for 'method=function_calling' will
             # always be an AnthropicTool
             tool_name = formatted_tool["name"]
-            if self.thinking is not None and self.thinking.get("type") in (
-                "enabled",
-                "adaptive",
-            ):
+            if (
+                self.thinking is not None
+                and self.thinking.get("type") in ("enabled", "adaptive")
+            ) or not _supports_forced_tool_choice(self.model):
                 llm = self._get_llm_for_structured_output_when_thinking_is_enabled(
                     schema,
                     formatted_tool,
diff --git a/libs/partners/anthropic/tests/unit_tests/test_chat_models.py b/libs/partners/anthropic/tests/unit_tests/test_chat_models.py
index be9ca01d4b..3a9d67905c 100644
--- a/libs/partners/anthropic/tests/unit_tests/test_chat_models.py
+++ b/libs/partners/anthropic/tests/unit_tests/test_chat_models.py
@@ -2707,6 +2707,59 @@ def test_with_structured_output_root_combinator_raises_when_thinking_enabled() -
         chat_model.with_structured_output(_Either, method="function_calling")
 
 
+class _Person(BaseModel):
+    name: str
+
+
+_ANTHROPIC_TOOL_SCHEMA = {
+    "name": "_Person",
+    "input_schema": {"type": "object", "properties": {"name": {"type": "string"}}},
+}
+
+
+@pytest.mark.parametrize("model", ["claude-opus-5-5", "claude-fable-5-1"])
+@pytest.mark.parametrize("schema", [_Person, _ANTHROPIC_TOOL_SCHEMA])
+@pytest.mark.parametrize("thinking", [None, {"type": "adaptive"}])
+def test_with_structured_output_skips_forced_tool_choice_when_unsupported(
+    model: str,
+    schema: type[BaseModel] | dict[str, Any],
+    thinking: dict[str, Any] | None,
+) -> None:
+    """Models that reject forced `tool_choice` bind the tool without forcing it."""
+    chat_model = ChatAnthropic(  # type: ignore[call-arg, call-arg]
+        model=model,
+        anthropic_api_key="secret-api-key",
+        thinking=thinking,
+    )
+
+    with pytest.warns(UserWarning, match="method='json_schema'"):
+        structured = chat_model.with_structured_output(schema)
+
+    bound = cast("RunnableBinding", structured.first)  # type: ignore[attr-defined]
+    assert [t["name"] for t in bound.kwargs["tools"]] == ["_Person"]
+    assert "tool_choice" not in bound.kwargs
+    assert "output_config" not in bound.kwargs
+
+
+def test_with_structured_output_forces_tool_choice_when_supported() -> None:
+    """Models that accept forced `tool_choice` keep `function_calling`."""
+    chat_model = ChatAnthropic(  # type: ignore[call-arg, call-arg]
+        model="claude-opus-5",
+        anthropic_api_key="secret-api-key",
+    )
+
+    class Person(BaseModel):
+        name: str
+
+    with warnings.catch_warnings():
+        warnings.simplefilter("error")
+        structured = chat_model.with_structured_output(Person)
+
+    bound = cast("RunnableBinding", structured.first)  # type: ignore[attr-defined]
+    assert bound.kwargs["tool_choice"] == {"type": "tool", "name": "Person"}
+    assert "output_config" not in bound.kwargs
+
+
 def test_get_num_tokens_from_messages_filters_unsupported_tools() -> None:
     """Token counting and sending agree on which tools the API will accept."""
     chat_model = ChatAnthropic(  # type: ignore[call-arg, call-arg]