OpenAI's CEO admits ChatGPT got messy and wants to fix it
Tech giants are all-in on AI these days – I mean, it's hard to find a company that isn't throwing the term around. And the one that arguably kicked off the craze, the creator of ChatGPT, OpenAI, is now making an effort to be more transparent about what's next.
CEO Sam Altman just shared a roadmap on X, revealing that OpenAI's upcoming GPT-5 release will feature its o3 reasoning model and, notably, be available to free users. He also mentioned that the company is working on making ChatGPT even easier to use.
Right now, if you want ChatGPT to use a specific model, you have to pick one from the model selector that pops up after the chatbot responds. If you're on the Plus or Pro plan, that menu can get pretty cluttered with different models and intelligence settings – which I guess could get pretty frustrating if you're shelling out $200 a month for the Pro tier.
As for OpenAI's roadmap, Altman revealed that GPT-4.5 will be the last model without "chain-of-thought" reasoning. That means every model after it will be able to break down problems into smaller steps to improve reasoning.
Once GPT-4.5 is out, OpenAI's main focus will be unifying its o-series and GPT-series models, creating a system that can seamlessly switch between tools, decide when to take its time thinking, and generally handle a wide range of tasks more effectively.
As for when all of this is happening, Altman didn't give a firm date for either GPT-4.5 or GPT-5, only mentioning that they could arrive in the coming weeks or months.
CEO Sam Altman just shared a roadmap on X, revealing that OpenAI's upcoming GPT-5 release will feature its o3 reasoning model and, notably, be available to free users. He also mentioned that the company is working on making ChatGPT even easier to use.
We want AI to "just work" for you; we realize how complicated our model and product offerings have gotten. We hate the model picker as much as you do and want to return to magic unified intelligence.
– Sam Altman, OpenAI CEO, February 2025
Right now, if you want ChatGPT to use a specific model, you have to pick one from the model selector that pops up after the chatbot responds. If you're on the Plus or Pro plan, that menu can get pretty cluttered with different models and intelligence settings – which I guess could get pretty frustrating if you're shelling out $200 a month for the Pro tier.
As for OpenAI's roadmap, Altman revealed that GPT-4.5 will be the last model without "chain-of-thought" reasoning. That means every model after it will be able to break down problems into smaller steps to improve reasoning.
Once GPT-4.5 is out, OpenAI's main focus will be unifying its o-series and GPT-series models, creating a system that can seamlessly switch between tools, decide when to take its time thinking, and generally handle a wide range of tasks more effectively.
For GPT-5, the company is planning to roll it out as a system that brings together a bunch of the company's latest tech, including the recently introduced Deep Research feature.
When GPT-5 does land, OpenAI intends to offer free users unlimited access – though with restrictions to prevent abuse – at the "standard intelligence setting." Plus, subscribers will get access to a more advanced version, while Pro users will be able to tap into an even higher intelligence tier.
As for when all of this is happening, Altman didn't give a firm date for either GPT-4.5 or GPT-5, only mentioning that they could arrive in the coming weeks or months.
FCC OKs Cingular\'s purchase of AT&T Wireless
Things that are NOT allowed: