When Will ChatGPT-5 Be Released (August 2025 Update)
OpenAI has released GPT-4.5, the last model before ChatGPT-5. Clues from Sam Altman's social media posts suggest that GPT-5 might be just around the corner.
OpenAI, the company behind ChatGPT, hasn’t publicly announced a release date for GPT-5.
But we've been watching developments closely.
- February 27, 2025: GPT-4.5 "Orion" was released to ChatGPT Pro users
- June 18: Sam Altman, OpenAI's CEO, confirmed the GPT-5 release date would be summer 2025
- July 12: OpenAI's open-source model was indefinitely delayed. It's now claimed it will be released before GPT-5
- July 19: The "gpt-5-reasoning-alpha-2025-07-13" model was reported on X
- July 24: The Verge reported a new release date for GPT-5: early August 2025
- July 24: On Windows, users began to report a new "Smart Mode" in Copilot
- July 29: ChatGPT users reported a GPT-5 rollout alongside a video chat feature
- July 30: "gpt-5-auto" and "gpt-5-reasoning" reportedly added to MacOS app configuration settings
- August 2: GPT-5 reported to outperform o3-mini-high in a lateral reasoning benchmark
In this article, we’ll analyze these clues to estimate when ChatGPT-5 will be released. We’ll also discuss just how much more powerful the new AI tool will be compared to previous versions.
ChatGPT-5 Release Date
According to the latest available information, ChatGPT-5 is due to be released in early August 2025 with mini and nano versions.
The most recent news comes from an August 2nd post on X. A user with early access to GPT-5 reported that gpt-5-reasoning outperforms o3-mini-high in a new lateral reasoning benchmark.
GPT-5 also appears to outperform proprietary models like Claude 4 Opus and Gemini 2.5 Pro Thinking.
Prior to this, Sam Altman made a statement in a video podcast on the official OpenAI YouTube channel on June 18th, 2025, saying the release would happen "this summer."
This confirms previous statements that we've been tracking,
An August release now looks inevitable, based on the timeline we have.
In July 2025, OpenAI delayed the release of its open-source model.
The eagerly anticipated open weights reasoning model was pushed back in June, then pushed back again for an indefinite amount of time on July 12.
Altman said: "We are delaying [the open weights model]; we need time to run additional safety tests and review high-risk areas. We are not yet sure how long it will take us."
Tom Warren at The Verge expanded on this in late July. He said that, "OpenAI is trying to ship it before the end of July — ahead of GPT-5’s release. Sources describe the model as “similar to o3 mini,” complete with reasoning capabilities."
In April, he said ChatGPT-5 was set to be released “in a few months”.
At the same time, he announced the plans to release both an o3 and o4-mini model in “a couple of weeks.”
Both of these top large language models were released on April 16.
In February, he said the release date for GPT-5 was set for "weeks/ months" from the date of an announcement on X.
The statement notes that GPT-4.5 would be released "next", with GPT-5 following sometime later in the year.
GPT-4.5 was indeed released shortly afterward on February 27th for Pro users. Plus and Team users got access in March.
Also in February, Stephanie Palazzolo reported on X that OpenAI leaders told employees GPT-4.5 was imminent. And Pro users saw GPT-4.5 Research Preview in the beta app.
We will likely see similar leaks before GPT-5 is released.
Sam Altman's February tweet confirmed that GPT-4.5 is the chain of thought model that was previously referred to as "Orion".
OPENAI ROADMAP UPDATE FOR GPT-4.5 and GPT-5:
— Sam Altman (@sama) February 12, 2025
We want to do a better job of sharing our intended roadmap, and a much better job simplifying our product offerings.
We want AI to “just work” for you; we realize how complicated our model and product offerings have gotten.
We hate…
GPT-5 will bring together various models, including o3, which will no longer be available to use when GPT-5 is available.
However, the existence of GPT-5 had already been all but confirmed months prior.
In September 2023, Sam Altman gave a speech at his former venture capital firm, Y Combinator. He publicly confirmed that GPT-5 and GPT-6 were “in the bag.”
A few months later, in November 2023, he told the Financial Times that OpenAI was actively working on GPT-5.
In January 2024, Altman gave an interview on Bill Gates’ podcast, in which he confirmed that OpenAI was actively developing GPT-5.
And in October 2024, a source claimed that "Orion" was due for release in December 2024. Sam Altman posted that the claim was "fake news".
Despite these confirmations that ChatGPT-5 is, in fact, being created, OpenAI has yet to announce an official release date.
Why GPT-5 Might Be Released Soon
Altman's post on X, which was shared in February 2025, confirms that GPT-5 was still on the roadmap. And it confirms that OpenAI is changing the way it provides large language models to its users.
In the ChatGPT interface, users have to choose the model they want to use depending on the task they want to work on. To eliminate confusion, GPT-5 will replace the model switcher entirely.
Altman also confirms that users on the free ChatGPT plan will get unlimited usage of GPT-5 for free. Anyone who pays for access will be able to run it "at a higher level of intelligence".
"Smart Mode" in Windows Copilot appears to be a new GPT-5 mode for Windows users.
Pressure From Competing Models May Speed Up the Release
While ChatGPT was revolutionary on its launch a few years ago, it’s now just one of several powerful AI tools.
With competitors pouring billions of dollars into AI research, development, and marketing, OpenAI needs to ensure it remains competitive in the AI arms race.
As we waited for GPT-5, companies developing competing LLMs have had time to catch up.
Google is pushing ahead with the launch of AI Mode, which uses a chatbot-like experience in Google search.
After rolling out in the US, AI Mode is now available for English-language queries in India from June 24th, 2025, and the UK from July 28th.
Google has also increased the number of AI Overviews showing on organic search results pages. Now, more than 13% of regular Google searchers trigger an AIO.
This has left businesses looking for data on AI search habits. Tools like Semrush Enterprise AIO provide brand-new tracking tools for several AI search platforms, including ChatGPT.
Released in January 2025, DeepSeek R1 is a Chinese model that replicates the reasoning capability and performance of OpenAI's o1. Importantly, DeepSeek is open source and free.
DeepSeek costs a fraction of the amount to train. This shook the stock market and introduced some new doubts over OpenAI's potential to be profitable.
On February 24th, 2025, Anthropic released Claude 3.7, its first hybrid reasoning model. Anthropic said that Claude 3.7 has been developed for real-world business usage rather than specialized tasks, like complex math problems. It has since released Claude 4.
Since GPT-5 will merge several models that already exist, it's reasonable to expect that it will follow soon after GPT-4.5 "Orion". The question is: how long will it take for GPT-4.5 to be released?
ChatGPT-4 was released just three months after GPT-3.5, so users are accustomed to seeing regular model releases from OpenAI.
Time Since Previous ChatGPT Release
| Model | Release Date | Months Since Previous Release |
| GPT-2 | February 14, 2019 | |
| GPT-3 | June 11, 2020 | 15 |
| GPT-3.5 | November 30, 2022 | 29 |
| GPT-4 | March 14, 2023 | 3 |
| GPT-4o | May 13, 2024 | 13 |
| GPT-o1 | September 12, 2024 | 3 |
| GPT-o3 | January 31, 2025 | 4 |
| GPT-4.5 | February 27, 2025 | 2 |
| GPT-4.1 | April 14, 2025 | <1 |
| GPT-o3 | April 16, 2025 | <1 |
| GPT-o4-mini | April 16, 2025 | <1 |
| GPT-5 | Expected in August 2025 |
OpenAI's releases have been faster in the past 12 months. But seems there have been delays in bringing GPT-5 to market.
In March 2024, Sam Altman was interviewed on the Lex Fridman Podcast. He told Lex that OpenAI plans to release an “amazing new model this year.”
Additionally, Business Insider published a report about the release of GPT-5 around the same time as Altman’s interview with Lex Fridman. Sources told Business Insider that GPT-5 would be released during the summer of 2024.
Altman could have been referring to GPT-4o, which was released a couple of months later. However, it’s also possible that he meant GPT-5.
OpenAI is facing the same challenge that other LLM providers face: a lack of training data. The solution is to use synthetic data: information created for the specific purpose of training LLMs.
If OpenAI has resolved its training challenges by using synthetic data, then GPT-5 will be released this year.
According to Business Insider, certain enterprise customers — organizations subscribed to ChatGPT Enterprise, the highest tier currently available — have been granted early access to demo versions of GPT-5.
However, the new model still needs to complete its training. It also needs to be safety tested. Both of these processes could significantly delay the release date.
Why GPT-5 Is Later Than Expected
Safety testing is likely the largest hurdle facing GPT-5’s release.
ChatGPT (and AI tools in general) have generated significant controversy for their potential implications for customer privacy and corporate safety.
OpenAI has already incorporated several features to improve the safety of ChatGPT. For example, independent cybersecurity analysts conduct ongoing security audits of the tool. They attempt to find vulnerabilities, which are then patched.
OpenAI also runs a “bug bounty” program, which rewards people for finding and reporting security flaws in ChatGPT.
However, many still have safety concerns about AI.
In March 2023, for example, Italy banned ChatGPT, citing how the tool collected personal data and did not verify user age during registration. The following month, Italy recognized that OpenAI had resolved the problems and allowed it to resume ChatGPT service in the country.
In May 2023, Sam Altman testified before the Senate Judiciary Committee in favor of government regulation to “mitigate the risks” of AI tools. Altman recognized the widespread fears about the potential dangers of AI, including misuse, and argued in favor of an agency with the power to license AI tools — and “take that license away” to “ensure compliance with safety standards.”
More generally, hallucinations remain a significant barrier. The BBC reported that AI chatbots are unable to produce accurate summaries of news articles, leading to concerns about misinformation online.
In July 2025, Sam Altman was quoted saying he had concerns about the extent of ChatGPT use among young people, particularly for therapy.
Altman said, "Right now, if you talk to a therapist or a lawyer or a doctor about those problems, there’s legal privilege for it. There’s doctor-patient confidentiality, there’s legal confidentiality, whatever. And we haven’t figured that out yet for when you talk to ChatGPT."
It would be catastrophic for OpenAI to press for stringent safety standards and then fall short itself.
Therefore, it’s likely that the safety testing for GPT-5 will be rigorous.
Altman said as much to Bloomberg earlier this year. He indicated that OpenAI will “take its time” to ensure that GPT-5 is something it feels “good about and responsible about.”
On May 28, 2024, the same day that OpenAI formally announced the successor to GPT-4, the company also announced the formation of a Safety and Security Committee. The OpenAI Board tasked the committee with “making recommendations to the full board on critical safety and security decisions for OpenAI projects and operations.”
In September 2024, OpenAI released the committee's recommendations across 5 areas.
GPT-5’s own safety testing could take months.
Reports indicate that the model is still in the training phase. After it completes its training, it will be extensively tested for security. That involves the model being “red teamed,” a process in which internal and external testers attempt to find flaws and vulnerabilities in the model, which can then be addressed.
On the one hand, this testing might not bring up any major issues.
On the other hand, there’s really no limit to the number of issues that safety testing could expose. Delays necessitated by patching vulnerabilities and other security issues could push the release of GPT-5 well into the second half of 2025.
The uncertainty of this process is likely why OpenAI has so far refused to commit to a release date for GPT-5, beyond a vague promise that we'll get to use it in the summer of 2025.
ChatGPT-5 Features
Altman and OpenAI have also been vague about what exactly ChatGPT-5 will be able to do. However, Altman's posts on X give us some more clues.
What is the Capability of GPT-5?
“It’s not like this model is going to get a little bit better,” Altman said at the World Governments Summit in Dubai in February 2024. “It’s going to be better across the board.”
One year later, Sam Altman confirmed that GPT-5 will "incorporate voice, canvas, search, Deep Research, and more".
Right now, Plus, Team, Education, and Enterprise ChatGPT users can run 10 Deep Research queries per month. Users of the $200/ month Pro plan get 120 queries per month.
Generally, then, GPT-5 should be significantly "smarter" than its predecessors.
In practice, that could mean better contextual understanding, which in turn means responses that are more relevant to the question and the overall conversation. It also means fewer hallucinations.
Smarter also means improvements to the architecture of neural networks behind ChatGPT. In turn, that means a tool that can process data faster and more efficiently
GPT-5 will have multiple levels of intelligence: a standard level for free users and higher levels of intelligence for paying subscribers.
GPT-5 will replace the model switcher.
Sam Altman's tweet confirms that OpenAI feels its products have become too complicated. The switch in naming convention from 4o to o1 could be part of the problem.
In its announcement for o1, a model originally code-named "Strawberry", OpenAI said it chose the name o1 to "reset the counter back to 1" to reflect a new type of model. However, it led to friction in the ChatGPT interface as the model picker became more complicated for users.
With GPT-5, all of its o-series and GPT-series models will be combined into one product, and the model picker will disappear. Altman says that GPT-5 will "know when to think for a long time or not, and generally be useful for a very wide range of tasks".
Build a winning strategy
Get a complete view of your competitors to anticipate trends and lead your market
Smarter could also mean more extensive training.
OpenAI is training GPT-5 on vast amounts of data.
Most of that training data is scraped from the internet. But a significant proportion of its training data is proprietary — that is, purchased or otherwise acquired from organizations.
These proprietary datasets could cover specific areas that are relatively absent from the publicly available data taken from the internet. Specialized knowledge areas, specific complex scenarios, under-resourced languages, and long conversations are all examples of things that could be targeted by using appropriate proprietary data.
In theory, this additional training should grant GPT-5 better knowledge of complex or niche topics. It will hopefully also improve ChatGPT’s abilities in languages other than English.
GPT-5 is also expected to be more customizable than previous versions.
Individuals and organizations will hopefully be able to better personalize the AI tool to improve how it performs for specific tasks.
Smarter also means multimodal. GPT-4 is partially multimodal; it is able to process images. More recent updates have allowed ChatGPT to process audio, too. These updates “had a much stronger response than we expected,” Altman told Bill Gates. “We’ll be able to push that much further” in GPT-5.
Altman's 2025 announcement confirms that GPT-5 will be able to handle speech, images, reasoning, and research without the need to switch models.
All this “is a bigger deal than it sounds,” Altman claimed at the World Government’s Summit in Dubai. “What makes these models so magical is that they’re general,” he explained. “If it’s a little bit smarter, it’s a little better at everything.”
Will ChatGPT-5 Be Sentient?
It’s almost certain that GPT-5 will be smarter than ChatGPT-4 and ChatGPT-4o. If Altman’s promises come true, it will be better at everything.
But does this mean just another incremental improvement to AI tools, or does it mean the arrival of AGI?
AGI, or artificial general intelligence, is the concept of machine intelligence on par with human cognition. A robot with AGI would be able to undertake many tasks with abilities equal to or better than those of a human.
A famous scenario used to demonstrate — or perhaps test for — the presence of AGI is called the Coffee Test. It was created by Steve Wozniak, the co-founder of Apple. The Coffee Test goes as follows: Imagine standing in your kitchen, and seeing a robot strolling in and making a cup of coffee. The robot was not programmed to walk into your kitchen, find the coffee maker, locate the correct ingredients, and brew a great cup of coffee. Yet it still decided to do so, and managed it.
The prospect of AGI is both exciting and terrifying. Major tech leaders, including Elon Musk, have warned that it is not far off. In March 2023, thousands of experts signed an open letter calling on AI labs to “immediately pause for at least 6 months the training of AI systems more powerful than GPT-4.”
A few months after this letter, OpenAI announced that it would not train a successor to GPT-4. This was part of what prompted a much-publicized battle between the OpenAI Board and Sam Altman later in 2023. Altman, who wanted to keep developing AI tools despite widespread safety concerns, eventually won that power struggle.
Despite all this worry about AGI, and predictions that it will arrive soon, GPT-5 is unlikely to make AGI a reality.
The model may be able to act more autonomously than previous models. For example, businesses might be able to build autonomous AI agents that could schedule events, respond to emails, and make purchases without direct human oversight.
In this sense, OpenAI has claimed that GPT-5 could “bridge the gap between human and machine communication” through its improved capabilities and better conversational experience.
But GPT-5 almost certainly won’t be sentient.
How Will ChatGPT-5 Be Different From Previous Models?
Nevertheless, GPT-5 will certainly be a great deal more capable than GPT-4. No one has been clearer about this than Sam Altman.
In March 2024, Altman told Lex Fridman that GPT-4 “kind of sucks.” He said that “the delta between 5 and 4 will be the same as between 4 and 3… It is our job to live a few years in the future and remember that the tools we have now are going to kind of suck looking backward at them.”
The Difference Between GPT-5 and GPT-4
We’ve actually covered many of the differences between GPT-5 and GPT-4 already.
The big leap forward is likely to be the merging of all GPT and o-series models into one. That means the model will choose how to process each prompt, rather than the user having to do it.
Multimodality is perhaps the most flashy difference. Improvements to customization, speed, power, and efficiency will all also be noticeable.
One improvement Altman has highlighted is reliability. AI tools, including the most powerful versions of ChatGPT, still have a tendency to hallucinate. They can get facts incorrect and even invent things seemingly out of thin air, especially when working in languages other than English.
“If you ask GPT-4 most questions 10,000 times,” Altman explained to Bill Gates, “one of those 10,000 is probably pretty good, but it doesn’t always know which one — and you’d like to get the best response of 10,000 each time.”
Altman implied that GPT-5 will be better able to identify that great response. That is, GPT-5 will be more reliable than GPT-4.
Additionally, GPT-5 will have far more powerful reasoning abilities than GPT-4. Currently, Altman explained to Gates, “GPT-4 can reason in only extremely limited ways.” GPT-5’s improved reasoning ability could make it better able to respond to complex queries and hold longer conversations.
In large part, that’s because GPT-5 is being trained on a much larger dataset than GPT-4.
| Model | Training Time (Years) | Retail Cost |
| GPT-3 | 405 | $9,000,000 |
| GPT-4 | 6,507 | $224,000,000 |
| Gemini | 15,616 | $440,000,000 |
| GPT-5 | 16,438 | $612,000,000 |
| Grok 3 | 32,855 | $1,200,000,000 |
Source: LifeArchitect
GPT-4 is Already Much Smarter Than GPT-3
As Altman noted, GPT-4 was already a significant improvement over GPT-3.
For starters, GPT-3 is unimodal. It can only process text inputs. Additionally, it was trained on a much lower volume of data than GPT-4. That means lesser reasoning abilities, more difficulties with complex topics, and other similar disadvantages.
GPT-3 also has far fewer parameters than GPT-4. Parameters essentially determine how an AI model behaves. The number and quality of the parameters guiding an AI tool’s behavior are therefore vital in determining how capable that AI tool will perform.
In general, more parameters mean a more powerful AI. GPT-3 had roughly 175 billion parameters. While the number of parameters in GPT-4 has not officially been released, estimates have ranged from 1.5 to 1.8 trillion.
What Comes Next for OpenAI After ChatGPT-5?
Looking ahead, OpenAI’s internal roadmap, partially revealed in this heavily redacted leaked memo, suggests the next evolution of ChatGPT will be far more than an incremental update. Instead, the company is preparing to unveil what it refers to internally as a “super-assistant,” powered by the upcoming o3‑pro model, which brings deeper reasoning, autonomy, and software-level control.
From Chatbot to Entity
According to the memo, OpenAI plans to position ChatGPT as a T‑shaped user entity broadly capable of everyday tasks. These would include managing calendars, booking travel, sending emails, while also offering deep expertise in complex domains such as programming or strategic research. This combination would allow ChatGPT to act as a trustworthy and emotionally intelligent personal assistant rather than a simple chatbot.
Tools That Take Action
Crucially, the new assistant is described as having agentic abilities. Features such as “Computer Use” are designed so ChatGPT can interact with applications and interfaces on behalf of the user, taking real-world actions rather than simply offering advice.
Search That Lives Inside It
OpenAI is also reportedly building a proprietary search index, allowing the assistant to navigate the web and take action without relying on traditional search engines . This move would reinforce ChatGPT’s role as an active interface, not just a passive chatbot.
Timing & Preview Signals
The memo indicates that the transition to this super assistant mode was targeted for the first half of 2025. With a limited preview of o3‑pro already underway among select users, OpenAI is expected to announce the full public rollout very soon.
What This Means for Users
-
Expect ChatGPT to move from answering questions to performing tasks proactively.
-
Think less “what can ChatGPT answer?” and more “what can it do for me?”
-
This next release reportedly combines broad usability and deep expertise, making the experience feel more personal, proactive, and integrated.
Key Takeaways
Ultimately, until OpenAI officially announces a release date for ChatGPT-5, we can only estimate when this new model will be made public.
There are a number of reasons to believe it is imminent, and GPT-5 will likely launch in August 2025.
Stop Guessing, Start Growing 🚀
Use real-time topic data to create content that resonates and brings results.
Exploding Topics is owned by Semrush. Our mission is to provide accurate data and expert insights on emerging trends. Unless otherwise noted, this page’s content was written by either an employee or a paid contractor of Semrush Inc.
Share
Newsletter Signup
By clicking “Subscribe” you agree to Semrush Privacy Policy and consent to Semrush using your contact data for newsletter purposes
Written By
Josh is the Co-Founder and CTO of Exploding Topics. Josh has led Exploding Topics product development from the first line of co... Read more