Replicating the UI deep research results using API

API
1

I have crafted a prompt that gives me optimal results via the UI when I enable Deep Research with O3 model. I would like to use the same prompt programmatically with Deep Research API. I see that there are a lot of tunable parameters. I’m just wondering if I can get the same parameters as the UI to get the same results.

Currently this is what I’m using:

  response = client.responses.create(
          model="o3-deep-research",
          input=[
              {"role": "developer", "content": [{"type": "input_text", "text": system_prompt}]},
              {"role": "user", "content": [{"type": "input_text", "text": user_prompt}]}
          ],
          tools=[
              {"type": "web_search_preview"},
              {"type": "code_interpreter", "container": {"type": "auto"}},
          ],
          background=True
      )

I’m wondering if I could access the official Deep Research system prompt that the UI version uses. Also if there are other parameters that I need to tweak: temperature, seed, etc…

1 Like
2

I think that the closest official thing you can get, is what is in the cookbook:

It recommends a preparatory approach to enrich the prompts before inquiring the deep research models.

As for the exact prompts, I believe it is not possible as ChatGPT is a closed product.

3

Thanks for chiming in. I had seen the cookbook post. I was wondering if someone from OpenAI could help me with the exact parameters.