Any experiences with chatGPT/ OPENAI's deep research?

[noname223]

I am very curious about that feature. It is not available in Germany yet. But I think US users can try it. Especially the ones who pay for chatGPT 20 bucks.
To make it short. The normal chatGPT usally is as good as an undergraduate student. but deep research is as good as a PhD candidate. For giving an answer it takes up to 30 minutes though.

I will add a link where it is described. It is limited how many deep research request you can do per month.

https://help.openai.com/en/articles/10500283-deep-research-faq

OpenAI Launches Deep Research (New AI Feature for ChatGPT): What it Do and How to Use It - GeeksforGeeks

Explore OpenAI's Deep Research for ChatGPT: Unleash AI-powered data analysis, real-time research, and smart citations. Learn step-by-step how to use this groundbreaking tool for academic, market, and technical insights.

www.geeksforgeeks.org

 

[GlassMoon]

It has been available in Germany via the subscription for some time already. I have the $20 subscription. I tested it on a scientific topic I'm firm in and it did a good job summarizing the latest results and citing the latest publications. You get a list of sources it has read. Afterwards, you can chat about the topic with the 4o model or switch to a reasoning model if needed. Seems useful to me! Reminds me that I wanted to ask it to research another topic for me :-)

The main difference to the 4o model is that it can do a multi-step research. It reads through the first set of results, then concludes where to go from there, and repeats.

 

[derpyderpins]

I just swapped my chatgpt for Gemini and grok as I make my rounds trying them all.

I had a big task load on a tight deadline from Friday to Monday, and honestly using those two made me get unthinkable amounts done.

Legal research has always been a glaring weakness with ai, and it seems the new deepsearch and reasoning models are finally turning that corner a bit.

The issue is always case law. They'll give you a quote and case citation, but you go to find that case yourself and it wouldn't exist. Didn't have that happen once this weekend, although the conclusions they drew weren't super reliable, I got a few pretty great leads/arguments that definitely saved time over traditional searching.

That combo is working well for me. Gemini for the Google docs integration, and grok is honestly performing better than chatgpt did in my experience. It's nice being able to switch from normal to reasoning to deepsearch in the same chat, which chatgpt couldn't do as of a month ago if you had attachments in the chat and one of the models wasn't compatible with the file type. Of course they're all always releasing new models then catching up. Inevitably I think the difference will come from integrations and functions.