
- Openai appears to test GPT-5 based on leaked files and internal biocuriti tools
- GPT-5 is rumored to complete memory, region, vision and task in one
- No specific release date has been announced, but it is likely to be officially rolled out in the next few months
Rumors that OpenaiI is already testing the GPT of 5, has begun to spread thanks to the leaks and the signals that appear online. On July 13, 2025, a particularly remarkable example on X was shared by Engineer Tibor Blaho on posting a partial screenshot of a configure file in “GPT, 5 Reasoning Alpha”.
The same week, independent researchers referred to GPT -5 in OpenAII’s internal biocecmark benchmark repository, suggesting that the model is already being tested in sensitive domains such as bio -safety.
If those indirect porters were not enough, then Zikun Zhang of Openai Clearly said GPT-5 “coming” during the discussion of the new Chatgpt agent facility.
The attention of the GPT-5 release date is not surprising because people are asking Openai about it as the company released GPT-4 in 2023, and was intensified only when GPT-4.5 came out last year. The Chatgpt agents certainly speculate on the fever pitch, as these digital assistants can go out and do it online for you such as booking tickets and organizing a calendar. There are all those things that people have thought that GPT-5 will bring in chat.
GPT-5-resoning-alpha-2025-07-13h/t @swishfever pic.twitter.com/hq5csd0ic6July 19, 2025
GPT-5 future
The response of people using Chatgpt agents can actually be used to complete training for GPT-5. And if rumors about Million-token reference window are true, it will require a lot of training. The same goes to the allegedly integrated nature of GPT-5.
This would mean that the GPT-5 will not include switch between features such as visual analysis, code interpretation and similar capabilities as chatgate agents; It will serve as a unique system. You can ask it to explain an image, send an email, schedule a meeting and create a vocal summary with the same sign and perform an outspoken summary.
For example, a parent can coordinate the school schedule, food plans and the final-minute birthday party logistics, or you can plan a trip, book hotels, place it in your calendar, and email your family with a single request. The GPT-5 is believed to have been designed to address issues of hallucinations and fine misconceptions, making it rely more than GPT-4 or GPT-4.5.
There is also a question of memory. Openai is quietly rolling long-term memory in the chatgpt, which means that the model misses things about you, but GPT-5 can make it even more powerful.
Of course, security questions are already bubbling. The mention of GPT-5 being tested in biocratic contexts has left people. If it is enough to help help in complex research about biology, can it also spit dangerous information when it is wrongly motivated? Openai has promised to manufacture in safety measures, but history tells us that people are very good in finding clever ways around the digital fence.
Regardless, we will all find out soon, although depending on the previous launch, it is likely to be limited to customers of high levels of Chatgpt at first. But you can bet that Openai will not release GPT-5 until they are sure that it would not embarrass them in the first day.

