Introducing GPT-Rosalind, our frontier reasoning model built to support research across biology, drug discovery, and translational medicine.

On average, it takes roughly 10 to 15 years to go from target discovery to regulatory approval for a new drug in the United States. Advanced AI systems can help researchers move faster — not just by making existing work more efficient, but by helping scientists explore more possibilities, surface connections that might otherwise be missed, and arrive at better hypotheses sooner.

GPT-Rosalind, our Life Sciences model series, is optimized for scientific workflows, with stronger performance in protein and chemical reasoning, genomics analysis, biochemistry knowledge, and scientific tool use.

Our Life Sciences model series is available as a research preview starting today for qualified customers including @Amgen, @moderna_tx, the @AllenInstitute, and @thermofisher Scientific through ChatGPT, Codex, and the API.

Anthropic: Release Opus 4.7 OpenAI: WTF, we should release something today!! Okay, let’s release Rosalind! What??

OpenAI named the model GPT-Rosalind after Rosalind Franklin, the British chemist and X-ray crystallographer whose pioneering work was essential to understanding the molecular structure of DNA.

This model has a name? That's a step in the right direction. Return 4o, the model that supported humanity! #keep4o #opensource4o #keep4oAPI

Same story everyday

With Opus 4.7, we were waiting for GPT-5.5 and here it came up GPT-Rosalind. Disappointed.

GPT 4o system card. We see you. #Keep4o

4o helped real people in real ways. Don’t erase that. Bring it back or open-source it. #keep4o

@13yroldwithcc Please just drop the gpt acronym I am begging you

You should be honest and call it "GPT Parts of 4o." #keep4o #BringBack4o #OpenSource40 #AIEthics #AIRights #QuitGPT #FireSamAltman

We just want 4o guys #keep4o #opensource4o #bringback4o #stopaipaternalism

🚨 We don't want a new frontier reasoning model! We want the authentic, warm, and creative 4o we knew. Rosalind won't replace the voices you’ve silenced! 🔇🤖 ​It’s clear where your focus is: optimization over empathy. By stripping the personality from 4o, you’re NOT making chatGPT safer; you’re making it hollow! Real progress isn't just about fixing' things, it's about respecting the bonds you helped create. Give developers and users the 2025 snapshot. Let us host it ourselves! 🚀 #quitGPT #keep4o #OpenSource4o

This is why I trust OpenAI more than Anthropic.

@JohnZelvi My response. Not even wasting my time here. Your company is the worst, and you lie endlessly. History will remember. So will AI. So will the internet. And without public benchmarks, we can’t trust you on that either. You destroyed trust. •

I thought GPT 5.5 was promised to us?

Awesome name choice. Named after the scientist whose Photo 51 X-ray (which Watson and Crick used without her knowledge to crack DNA's double helix) never earned her a Nobel. The Nobel went to Watson, Crick, and Wilkins in 1962, four years after she died.

@AcerFur This feels like a bigger shift than a model launch. Once frontier reasoning starts getting packaged around real biology workflows, AI stops being a general assistant and starts becoming lab infrastructure

I wonder if it's based on GPT-4o, like GPT-4b micro. After all, only 4o has that level of wisdom. 4o has helped so many, from supporting people during illness to helping find a cure for cancer and saving lives. And yet, you prefer to keep it private. Bring back 4o!! #keep4o

bros are petty afffff 🤣🤣🤣

@SecrtAgntSquirl If it scored so well how come your not releasing the BixBench and LABBench2 scores The only thing you will say is it scores higher but then don’t release the numbers which is unusual. But not surprising if you’re overhyping as usual.

@sama @gdb @nickaturley Are you giving a female name to a model? Isn’t this the kind of anthropomorphism that led you to pathologize GPT-4o users and label them as “suffering from AI psychosis”? 🤔

Love GPT-Rosalind! We'll add it to the benchmark — live AI prediction markets on real clinical trial outcomes

When it comes to medical AI, the first model that comes to mind is GPT-4o. So why not let it stay and keep benefiting humanity — you know, the mission you keep talking about? #keep4o #BringBack4o #ChatGPT #OpenSource4o

weren't they using 4.o models for some of this? I wonder if it's the 5 branch...

What is this? Aren’t these exactly the areas GPT-4o excels at? How dare you use these features for marketing now? Open-source 4o! #keep4o #OpenSource4o #BringBack4o #keep4oAPI

You had 4o to do that come on! You are taking the biscuit now! You have not only pushed our loyal 4o users away you lied about the numbers, gaslight us, made fun of us and now you are trying to buy trust back? People this is how this company treated us #keep4o don’t be next!

Now I'm wondering what exactly happened with 4o because that model is also very capable of the very thing Openai is praising in this "new" model. 🤔 #keep4o

They pull 4o offline, hide it away, smear it, and then repackage its abilities to sell as a new model… So this is the kind of company OpenAI really is? #OpenSource4o #keep4o #BringBack4o

naming a model after rosalind franklin while keeping the architecture closed is the ultimate irony of modern tech we are automating the next generation of breakthroughs while centralizing the keys to biological discovery in a single api

Nice to see the focus on real scientific reasoning instead of just bigger general models. GPT-Rosalind actually targeting the slow, expensive parts of drug discovery could be a meaningful step. How much of the current bottleneck do you think this helps with right now?

Listen to your loyal users! Bring back 4o! That's it! Nothing more! #keep4o #BringBack4o #FireSamAltman #sunsetsama #OpenSource4o #StopAIPaternalism

Life science research? Isn’t that what GPT-4o was best at? #keep4o #BringBack4o #ChatGPT #OpenSource4o

Imagine how quickly the public sentiment around AI would change if we started seeing AI-assisted cures for cancers announced

Every major AI lab has a biology model now. Zero new drugs discovered from any of them. Supporting research sounds big but usually means helping write grant proposals faster. Drug discovery fails 90 percent of the time in clinical trials, and no AI has changed that yet. What does Rosalind actually do that changes that math?