- OpenAI seems to be testing GPT-5, based on leaks and internal biosecurity tools.
- GPT-5 unifies memory, reasoning and vision into one.
- There is no official release date, but it is likely to roll out within the next few months.
Rumors have spread online that OpenAI is Tibor Blaho, an engineer at X, shared a screenshot of a config that hinted at “GPT-5 reasoning alpha” dated July 13,2025.
Independent researchers found a mention of GPT-5 within OpenAI’s BioSec Benchmark repository that same week. This suggests the model has already been tested in sensitive domains such as biosecurity.
If those indirect portents weren’t enough, OpenAI’s Xikun Zhao The new ChatGPT agent feature was discussed and it was “is coming” explicitly stated GPT-5 would be”is coming”in the discussion.
It is not surprising that people are asking OpenAI about GPT-5 since GPT-4 was released in 2023. The interest only increased when GPT 4.5 was released last year. ChatGPT Agents brought speculation to a fevered pitch. These digital assistants are capable of booking tickets, organizing calendars and other online tasks. All of these things were thought to be brought by GPT-5 to ChatGPT.
gpt-5-reasoning-alpha-2025-07-13h/t @swishfever pic.twitter.com/Hq5csd0iC6 July 19, 2025 (19659012)
GPT-5 Future
The feedback received from users of ChatGPT Agents could be used to complete training for GPT-5. If the rumors of the context window with a million tokens are true, then it will require a lot of training. GPT-5 is supposedly unified.
This would mean GPT-5 will not only include features like visual analysis and code interpretation; it will also have the same capabilities as ChatGPT Agent. It will operate as a single system. You could ask the system to perform multiple tasks from a single prompt, such as interpreting an image, sending an email, scheduling a meeting and composing and performing a vocal summary.
A parent could, for example, coordinate school schedules and birthday party logistics at the last minute, or you could plan and book hotels, add it to your calendar, then email your family with all the details. GPT-5 was allegedly designed to address issues of hallucination, nuanced misunderstanding and allow people to trust it better than GPT-4 or GPT 4.5.
Subscribe to the latest news, reviews, opinions, top tech deals and more.Then there’s the issue of memory. OpenAI has quietly rolled out long-term memories in ChatGPT. This means that the model remembers you, but GPT-5 could make it even stronger.
The safety questions are already bubbling. People are spooked by the mention of GPT-5 in biosecurity contexts. Could it also give out dangerous information, if it is prompted in the wrong way, if it can reason about biologically complex research? OpenAI has promised safeguards, but the past shows that people are adept at finding ways to circumvent digital fences.
We will soon find out, but based on past launches, it is likely that the launch will be limited at first to subscribers to ChatGPT’s higher tiers. You can bet OpenAI will not release GPT-5 before they are sure it won’t embarrass themselves on day one.
- OpenAI claims that the new ChatGPT agent will run your errands and build your slides and make you appear to have your life in order
- AI has made a huge leap and now a quarter Gen Z believes AI is conscious.
Eric Hal Schwartz has been a freelance writer at TechRadar for more than 15 years. He has covered the intersection of technology and the world. He was the head writer of Voicebot.ai for five years and was at the forefront of reporting on large language models and generative AI. Since then, he has become an expert in the products of generative AI, including OpenAI’s ChatGPT and Anthropic’s Claude. He also knows Google Gemini and all other synthetic media tools. His experience spans print, digital and broadcast media as well as live events. He’s now continuing to tell stories that people want to hear and need to know about the rapidly changing AI space and the impact it has on their lives. Eric is based out of New York City.
