We’ve rolled back last week's GPT-4o update in ChatGPT because it was overly flattering and agreeable. You now have access to an earlier version with more balanced behavior. More on what happened, why it matters, and how we’re addressing sycophancy: openai.com/index/sycophan…

AMA at 9:30am PT with Head of Model Behavior Joanne Jang to talk all things ChatGPT's personality. reddit.com/r/ChatGPT/comm…

we missed the mark, learned a lot, and wrote it up:

some thoughts on human-ai relationships and how we're approaching them at openai it's a long blog post -- tl;dr we build models to serve people first. as more people feel increasingly connected to ai, we’re prioritizing research into how this impacts their emotional well-being.

recently the NYT asked a court to force us to not delete any user chats. we think this was an inappropriate request that sets a bad precedent. we are appealing the decision. we will fight any demand that compromises our users' privacy; this is a core principle.

Super excited about this one! Seeing the feedback roll in has been great.

It’s hard to believe it’s been 20 years since Steve told Stanford graduates to stay hungry and stay foolish. His powerful advice still rings true, and I hope it helps guide this year’s new grads as they begin their journey to become tomorrow’s leaders. stevejobsarchive.com/exhibits/stay-…

It it better to give users the answer, or help them understand how to find answers themselves? A 'good' answer for the same user might depend on which 'mode' they're in - information seeking or learning. On ChatGPT, users can now use study mode for learning. Nice work, Study Mode

what’s that smell … no, it couldn’t be … is that the big model smell?

A few months ago I joined OpenAI as a PM in research to work on model behavior. I am motivated by the many loved ones in my life who use our models regularly to get work done, explore ideas, pursue new goals, and as a trusted thought partner & sounding board. My north star:

Thank you Junhua Mao for believing in this project & rallying a team around the idea!

you can try it out now by picking `robot` in settings -> personalization -> customize chatgpt. let us know what you think! we wanted to iterate a bit more with feedback, which is why it's a preview + a bit buried right now in the ui. all personalities (including improvements

GPT-5 has 4 new chat personalities: Cynic, Robot, Listener, Nerd. Find them in Customize ChatGPT in settings. Research preview. Text-only. Opt-in. Change anytime. Written by Robot in GPT-5.

GPT-5 (medium reasoning) is the new leader on the Short Story Creative Writing benchmark! GPT-5 mini (medium reasoning) is much better than o4-mini (medium reasoning). Claude Opus 4.1 shows gains over Opus 4.

Some of you asked me about my blind test, so I created a quick website for yall to test 4o against 5 yourself. Both have the same system message to give short outputs without formatting because else its too easy to see which one is which. gptblindvoting.vercel.app

Does anyone else use sleep to better their skills or think through hard problems? As a ski racer, I learned to dream of my technique to lock-in skills when awake, later I taught myself to dream in Japanese as I learned it, I can use it to solve tricky SQL problems that I couldn't

Glad to get this update to the Spec out! Check it out if you're curious about our latest thinking on the topics Jason mentioned.

Congrats on your first ship, Rae Lasko! Welcome to the team.