The Latest ChatGPT-5 Rumors —and ChatGPT Predictions
July 28, 2025
A few notes from the intersection of branding and AI.
More Certainty About an August Release for ChatGPT-5 (7/28)
Axios, the Verge, and others are reporting that GPT-5 will definitely launch in August 2025, likely early August. Read my earlier posts on the subject, below, for thoughts on some of the features and details. FYI: They may release an “open weights” model prior to that, but it won’t be the GPT-5 we’re waiting for. It’s still anticipated that the new model from OpenAI will be multimodal (text, image, video) as well as a combination of their current suite of models. Kind of an über model that’s both better and combines more features than any of their current models.
This launch news comes at the same time that new rumors are surfacing about OpenAI’s strategic plan to, basically, become the internet. Not joking. Rather than Amazon, Google, Microsoft, YouTube, Meta, Apple, etc, there will only be OpenAI. That’s what some people are saying about their ambitions. I’m sure those other companies won’t sit around and ignore this. But upstarts have a way of disrupting in a way that established behemoths just can’t, and can’t necessarily stop. See Google 1998 or Meta 2006. Given OpenAI’s rumored plan, the next few years are going to be crazy.
A few predictions:
The new AI web browsers (if that’s what we’ll even call them) are going to transform the way we connect with digital information. Agents, auto tasks, smart assistants will all be part of it (reminds me of this Shel Silverstein poem and also this one). Perplexity’s Comet is already out. Browsers from OpenAI and Anthropic are on the way. Will a browser be part of GPT-5?
We may ultimately (maybe even sooner) start doing all our online shopping through an AI portal. Will that make Amazon simply a logistics and delivery company rather than an online marketplace?
OpenAI also has a rumored social media app on the way. It has the potential to completely take over that channel (though Zuckerberg will probably have something to say about it). An AI-powered social app also has the potential to be a thousand times more addictive and worse for us than the current generation of social apps already are. That would be terrible unless OpenAI (or other AI social apps) is much more conscientious than Meta has ever been.
ChatGPT-5 will accelerate new college graduate and white collar job loss. By a lot. It’s time to start your strategic career positioning now. However, Nathaniel Whittemore, host of the AI Daily Brief podcast, isn’t quite so pessimistic. In a recent episode he discussed some of the ways Jensen Huang, CEO of Nvidia, believes that job growth will boom (let’s hope so!). Huang also has an interesting twin-factory theory.
Quick ChatGPT-5 Update (7/22)
What’s the delay?
The tech news has continued to buzz about the imminent (everything thinks) release of ChatGPT-5. Why? Despite OpenAI’s insane valuation, it’s been more than two years since they released ChatGPT 4.0. (And the consensus is that while GPT-4.5 was better, it’s really not mind-bendingly better than 4.0. As of May, they now default to GPT-4o, which is a multimodal model—makes images as well as text—that stands for “omni” and force you to click on “more models” to even find 4.5 anymore.) That’s a really really long time in AI years. GPT-4.0 is essentially a great grandfather type model, and it’s not entirely clear how much better GPT-4o is other than the image gen part. Why has it taken so long, and what’s the currently release timing?
Seems to me that OpenAI is fighting battles on a lot of fronts. First, their naming structure for models is a mess. Here’s a post I wrote on that back in May. And their naming has only gotten more confused, with o1, o3, o4-mini, o4-mini-high, 4.5, 4.1, and 4.1-mini. I follow the AI trends and have a hard time explaining all the differences. (They need a good copywriter!) How will GPT-5 help with that? Will it be named 5.0 or, because it’s a “reasoning model,” named like the other “o” products and become o5?? Or something like GPT-Next? Inquiring minds want to know! Well, if rumors are to be believed, GPT-5 will integrate features from o3 and 4o and other models, perhaps obviating the smorgasbord nature of their current naming conventions. This would be good!
Second, there have been rumors that GPT-5 is SO good that it’s taking much longer to test it properly. Specifically, some people fear that bad actors could use the new model to create novel bio-weapons. A good reason to delay launch!
Third, there are new “open source” models being launched out of China that have incredible capabilities and are either free or very low cost. OpenAI surely doesn’t want to launch a new model that’s already lagging behind other models.
Fourth, Agents. OpenAI just released a feature it calls Agent to go with a recent feature it called Operator. Agent can take your goal, break it into discrete tasks, and take actions (call an API, update a spreadsheet, send an email) that help you accomplish the task. Agents have been the AI talk of 2025, but so far it’s mostly been for enterprise audiences and the jury is not sending glowing reports back as many enterprise agents are not yet fulfilling the agent promise. Which means that agents that are really going to improve our individual or small business lives are still a ways out. OpenAI may be trying to change that with GPT-5. We’ll see. The first version may just be more personally “customizable.”
Fifth, the amount of AI hallucination actually went up when OpenAI released it’s o3 reasoning model. That’s a problem, especially if you want an agent to complete a task for you. At the moment, double checking to make sure everything’s right—your trip to Norway in June didn’t get booked in July to…Greece, for instance—could take as long or longer than just dealing with the task on your own. It’s like, theoretically your 9 year-old could do the dishes, but you probably have to check that it got done right, or at all.
ChatGPT-5.o Impact
The new model is supposed to be much faster. It’s also supposed to significantly impact industries like healthcare, education, and business, by transforming workflows and fostering human-AI collaboration. We’ll see about that, too.
People are both excited and apprehensive about the new model's potential to accelerate scientific discovery, personalize education, and improve efficiency. But there are concerns about bio-security, job displacement, and the further spread of misinformation.
So When’s It Coming?
Given that it’s late July, people are now thinking August or September. Lots of chatter about a model that looks like it could be GPT-5 being used on various benchmark tests. Some, like this Reddit user, say OpenAI will want to launch it’s updated open source o3 model ahead of the GPT-5 release, meaning extra time.
But it’s coming up on two and a half years since the seminal GPT-4.0 was released, so they’re undoubtedly under significant pressure. Not the least from the VCs who have spiked the OpenAI valuation.
Original June Post: People Are Talking About GPT-5 (7/02)
July 2025 Release?
There’s been a ton of speculation, for instance here, here, here, and here, this month about the imminent release of ChatGPT -5. Okay, I’m sure lots of that speculative hype has come from OpenAI about a coming product launch, but we’ll set that aside. The scuttlebutt has been that OpenAI wanted to call their last big model release 5, but it wasn’t as transformative as they had hoped and opted for 4.5 instead. That’s why people are so excited about their newest model coming in…well sometime in July or August, if the tea leaves can be believed. Read more tea leaves on this Hard Fork podcast that was released today.
During a June 18, 2025 podcast, Sam Altman, OpenAI CEO, said they expect to launch GPT-5 “this summer.” That would be between today and September 21, if you’re going to be pedantic about it. Multiple industry insiders are saying they expect a July or August release, pending the final safety check.
All that said, OpenAI has not officially announced a release.
Let’s Speculate on GPT-5!
But that doesn’t mean we can’t speculate about timing. Or what GPT-5 might be capable of. About GPT-5, Altman has said, “It’s not like this model is going to get a little bit better.” Here’s what people think is coming.
Unified Multimodal Abilities: GPT-5 will natively handle text, images, and voice in a single interface, so users won’t need to switch models or tools for different media types.
Real-Time Knowledge Integration: GPT-5 is expected to access more up-to-date information, reducing hallucinations and improving relevance for current events and business needs.
Smarter Task Execution: Enhanced contextual memory and logic will allow GPT-5 to handle multi-step workflows, automate tasks, and execute more complex instructions with minimal prompting.
Customizability: The model will be more adaptable, letting users and organizations personalize outputs for specific tasks, industries, or branding requirements.
Creative Tools: For branding and marketing your small business, you can anticipate better, more efficient brand storytelling abilities, including images, video, and podcasts without needing to switch modes or tools.
Auto Support: Customer support by way of AI chatbots will continue to improve. Competition on AI customer support will drive down the cost, and by mid-2026 there will be no reason not to use one for your business.
Tasks: The current Tasks feature should get a giant upgrade and become much easier to integrate into everyday work life.
Market Research: OpenAI’s current o3 reasoning model is really really good at long, well-researched reports, though it still suffers from hallucinations. The tech whisperers out there say that GPT-5 should get even better at market research while reducing the mistakes.
I’m looking forward to the rollout. We’ll start testing how it generates on-brand copy as soon as it’s released. More soon.
Getting Better at Your AI Game
I love The AI Daily Brief podcast. Here’s a recent episode the suggests 3 ways to get better at AI right now (while you’re waiting for GPT-5).
