Sam Altman Teases 'A Ton' Of ChatGPT Updates, But GPT-5 Might Be Delayed

GPT-5 has been the talk of the town lately, with reports claiming the highly consequential ChatGPT upgrade will drop in early August. Some reports suggested that GPT-5 would arrive after the launch of the highly anticipated open-source version of ChatGPT. OpenAI's open model was rumored to launch in the final week of July, after multiple delays.

July is now over, and there's still no open-source ChatGPT to download. These things happen. The biggest players in the field release new features on a regular basis. AI companies unveil new models and features every chance they get. The announcements come more frequently than we're used to in the tech world. We wait a year between iPhone launches, but the next-generation AI upgrade is seemingly always just weeks away.

AI companies like OpenAI release their latest innovations as soon as they're ready for mass consumption. ChatGPT Agent is one such example, a new AI model that can perform complex tasks on your behalf in a virtual computer of its own. The model arrived a few weeks ago, about half a year after OpenAI unveiled the Operator agent, now baked into ChatGPT Agent. The same goes for GPT-5 and the open-source ChatGPT model. OpenAI might need to conduct more safety tests before releasing the two models.

This is all speculation based on Sam Altman's most recent tweet. The OpenAI CEO posted a teaser on X over the weekend saying that the company has "a ton of stuff" to launch in the coming weeks. He also said that hiccups might be part of the schedule, suggesting potential delays to some of OpenAI's big launches.

"We have a ton of stuff to launch over the next couple of months–new models, products, features, and more," the CEO said, without revealing specifics. "Please bear with us through some probable hiccups and capacity crunches. Although it may be slightly choppy, we think you'll really love what we've created for you!" It's unclear why Altman felt the need to post this particular teaser on a Saturday. However, the tweet follows weeks of GPT-5 leaks and teasers, including Altman mentioning the ChatGPT upgrade on a few occasions.

"This morning, I was testing our new model, and I got a question. I got emailed a question that I didn't quite understand. And I put it in the model, this is GPT-5, and it answered it perfectly," Altman said during a podcast interview after a visit to Washington D.C. a few days ago. "And I really kind of sat back in my chair, and I was just like, 'Oh man, here it is moment' [...] I felt like useless relative to the AI in this thing that I felt like I should have been able to do, and I couldn't. It was really hard. But the AI just did it like that. It was a weird feeling."

During his visit, there were numerous GPT-5 teasers posted online from savvy AI users who discovered that unreleased ChatGPT versions were being tested online under various codenames. All these teasers suggested the GPT-5 release was imminent. After all, Altman promised the GPT-5 upgrade for ChatGPT several months ago. Even with delays, if there are any, GPT-5 should drop this year.

Interestingly, Altman shared the update a day after The Information published an article titled "Inside OpenAI's Rocky Path to GPT-5." The publication said that GPT-5 should launch as soon as August, echoing similar reports from The Verge and Axios. While it didn't say anything about a delay, that "rocky path" to GPT-5 hints at OpenAI's struggles to build a bigger, better AI experience.

This wouldn't be the first time we've heard about an AI company struggling to make meaningful upgrades to its models. It's not just about the lack of data to train the more advanced models or the increasing infrastructure costs. Some reports claim that next-gen models like GPT-5 might not deliver the same massive leap as the GPT-4 family.

That said, The Information also presented the same general set of GPT-5 features that have appeared in other reports. GPT-5 will do away with the model picker, automatically deciding how to assist the user. Some questions will get immediate replies, while others will require in-depth reasoning. Like its predecessor, GPT-5 will be a multimodal AI, able to interpret text, images, and audio. The context window should increase in GPT-5, which should lead to improvements in memory and give the AI the ability to handle more data.

The report also noted that while the model picker is going away, OpenAI will still offer users multiple GPT-5 experiences. A core GPT-5 version will be available on the web and in its apps. Then there's GPT-5 mini, which might work best for quick tasks. Finally, GPT-5 nano could be used on mobile devices and on other devices where an even lighter AI experience is needed.