o3-mini is not working with G-Eval due to wrong temperature and logprobs parameter

[RicardoBochnia]

Describe the bug
Using a custom G-Eval metric with o3-mini does not work at moment because the OpenAI API is called with temperature = 0.7 and logprobs = True.

To Reproduce
Using a custom G-Eval metric with o3-mini and try to measure it. The API call to OpenAI will fail.

Expected behavior
The temperature should be 1.0 and logprobs = False (or omitted entirely).

Current workaround:
g_eval.py
Not setting logprobs and top_logprobs self.model.a_generate_raw_response and self.model.generate_raw_response (top_logprobs is configurable already but not logprobs)

gpt_model.py
Setting the temperature to 1

        else:
           # check necessary, if model_name = o3-mini set temperature=1
            return ChatOpenAI(
                model_name=self.model_name,
                openai_api_key=self._openai_api_key,
                base_url=self.base_url,
                *self.args,
                **self.kwargs
            )

With these changes o3-mini (and also o1-mini if manually added to the list of supported models) works with deepeval.