OpenAI CEO Sam Altman Shares New GPT-5 Roadmap, Promises One AI to Rule Them All

Channeling his inner Steve Jobs, OpenAI CEO Sam Altman revealed plans on Wednesday to drastically simplify the company’s product lineup, merging its scattered collection of AI models into a single unified system.

The move comes just weeks after Chinese rival DeepSeek made waves by launching a simpler, more capable AI that threatened to overshadow OpenAI’s more complex offerings.

“We want to do a better job of sharing our intended roadmap and a much better job simplifying our product offerings,” Sam Altman wrote in a tweet.

OpenAI currently has a diverse (some might say confusing) model offering, with GPT being the creative LLM, “o” being the reasoning model, “Dall-e” being the image generator, “Sora” being a video generator, and “GPT-Vision” being a model for visual understanding.

Echoing Jobs’s famous catchphrase, Altman tweeted, “We want AI to ‘just work’ for you; we realize how complicated our model and product offerings have gotten.”

Briefly outlining the roadmap, Altman promised to revamp the confusing interface that forced users to choose between different AI versions for different tasks.

The solution? A return to what Altman called “magic unified intelligence” in which the company would merge its various text models, including the mysterious “o-series” and the GPT family, into a single system that could handle everything from casual chat to complex analysis.

“A top goal for us is to unify o-series models and GPT-series models by creating systems that can use all our tools, know when to think for a long time or not, and generally be useful for a very wide range of tasks,” Altman wrote.

The first step in that direction will be the release of GPT-4.5, known internally as “Orion.”

Expected to be coming within the following weeks and months as confirmed by Altman—it will be OpenAI’s last model built upon the traditional GPT technology, which is basically the ChatGPT we all know and love.

The real transformation will come with GPT-5, which is supposed to incorporate all of OpenAI’s technologies, including the capabilities of the unreleased o3 model.

In other words, GPT-5 will be a kind of AI hub capable of doing everything its various models do now, from creating images to creative writing to complex reasoning and programming, depending on the task. Users won’t need to choose which model is best for the task.

OpenAI’s so-called “reasoning models” work through a query via “chain-of-thought,” a technique in which a model analyzes a question and generates a sequence of intermediate reasoning steps before reaching a conclusion.

Typically, reasoning models are not good at creative tasks, and vice versa: models good at creativity (like GPT-4.5) tend to be bad at reasoning, especially when zero-shot prompting—users expecting to do everything in one single command—is involved.

No doubt provoked by its new-found competition with DeepSeek, whose surprise launch recently demonstrated its superior AI model available for free and open source, OpenAI announced today that GPT-5 will be available through ChatGPT’s free tier.

Users will get unlimited access to the standard intelligence setting, though usage will be monitored to prevent abuse.

For businesses and professional-grade research, the company unveiled a tiered access system that will give paying customers more computational muscle.

Plus, subscribers will access a higher intelligence setting that uses more computing power to solve complex tasks, while Pro users will get the most capable version.

“These models will incorporate voice, canvas, search, deep research, and more,” Altman wrote.

The changes also mean OpenAI will not release reasoning/coding models as a standalone offer—not even the currently available o3 model—instead incorporating its capabilities into the unified GPT-5 system.

For developers using OpenAI’s API, this translates into a simpler integration process—instead of choosing between different models, they’ll access a single system capable of handling diverse tasks.

However, it may also be counterproductive as specific use cases may be more profitable with less computing power, and less expensive price per token.

OpenAI didn’t elaborate on how it will impact API costs. But Sam Altman is sure it will be worth it—in fact, he expects it to be so capable that he hyped it up, saying, “I don’t think I’m going to be smarter than GPT-5” in a recent panel in Berlin.

As a famous man once said, “Stay hungry, stay foolish,” Sam.

Edited by Josh Quittner and Sebastian Sinclair