with_structured_output() does not work with Gemini 2.5 Models

[MeerrettichMeister]

Using structured output with the new Gemini models dont work.

from langchain_core.messages import HumanMessage
from langchain_oci import ChatOCIGenAI
from pydantic import BaseModel


class Formatted(BaseModel):
    answer: str
    user_request: str


oci_chat = ChatOCIGenAI(
    model_id="google.gemini-2.5-flash",
    service_endpoint="https://inference.generativeai.us-ashburn-1.oci.oraclecloud.com",
    compartment_id="compartment_id",
)


unstructured_response = oci_chat.invoke(
    [HumanMessage("Tell me something about the statue of liberty")]
)

structured_chat = oci_chat.with_structured_output(Formatted)
structured_response = structured_chat.invoke(
    [HumanMessage("Tell me something about the statue of liberty")]
)
print(unstructured_response)
print(structured_response)

Running the above code returns a proper response with the regular chat interface but fails with the structured output.
It either returns None or fails by the pydantic validation.

Python 3.12
langchain 0.3.27
langchain-oci 0.1.5
pydantic 2.11.7