
ZDNET's cardinal takeaways
- As nan overgarment connected GPT-5 dries, OpenAI is already moving connected GPT-6.
- The model's biggest differentiator seems to beryllium deeper personalization.
- OpenAI is keeping an progressively brisk gait connected nan way toward AGI.
Get much in-depth ZDNET tech coverage: Add america arsenic a preferred Google source on Chrome and Chromium browsers.
At nan opening of August, OpenAI launched its much-anticipated GPT-5 family of models. While nan motorboat was somewhat rocky, nan institution isn't letting that slow down its momentum: it is already teasing GPT-6, its adjacent procreation of models.
Also: Here are each nan GPT-5 updates OpenAI has rolled retired since launch
Last week, erstwhile chatting pinch reporters, OpenAI's Sam Altman shed immoderate ray connected GPT-6, sharing that it is not only already successful nan useful but that nan motorboat will travel sooner than nan hold betwixt GPT-4 and GPT-5, according to a CNBC report. Naturally, Altman besides teased immoderate upcoming features that would make GPT-6 a superior exemplary from its predecessors.
For a roundup of everything we know, return a look below.
What do we cognize astir nan model?
According to nan CNBC report, nan caller exemplary will usability much personally by responding to users and adapting to them. It will do truthful done an enhanced type of its Memory feature, which would retrieve a person's circumstantial preferences to amended cater to those needs.
Also: GPT-5 is friendlier now - but not everyone likes it. Here's why
"People want memory," Altman said to nan reporters. "People want merchandise features that require america to beryllium capable to understand them."
The type of nan Memory characteristic presently disposable successful ChatGPT was updated successful April to reference each of a user's past conversations to amended pass early responses (to move it on, toggle nan action accessible successful Settings nether nan Personalization tab). Despite saying nan updated representation characteristic was his favourite motorboat of nan past year, Altman added concerns astir privacy, arsenic nan impermanent representation is not encrypted. He said early encryption could beryllium a possibility, according to nan report.
Beyond personalization, Altman did stock that early versions of ChatGPT will person to comply pinch a recent executive bid from nan Trump administration that prevents "woke AI," which nan bid described arsenic models built pinch "ideological biases aliases societal agendas." He besides shared that nan institution is moving pinch psychologists to thief create early experiences for users.
When will it beryllium released?
While nary timeline has been shared yet, Altman did opportunity that nan spread betwixt GPT-5 and GPT-6 would beryllium smaller than nan 1 that followed GPT-4, which gives america a ballpark thought of erstwhile it will launch, if you return a look astatine nan anterior timeline.
Also: AI is creeping into nan Linux kernel - and charismatic argumentation is needed ASAP
When ChatGPT became celebrated successful November 2022, it was moving connected GPT-3.5. Shortly after, OpenAI released GPT-4 successful March 2023. OpenAI introduced GPT-4o -- a important update that improved its capabilities broadly and notably introduced multimodality -- a twelvemonth later, successful March 2024, but GPT-5 was not released until August 2025. This makes nan clip betwixt GPT-4 and GPT-5 complete 2 years, truthful astatine slightest you cognize it will beryllium sooner than then.
Even that very loose timeline is worthy taking pinch a atom of salt, arsenic merchandise motorboat timelines perpetually get pushed back. For example, successful February, Altman initially shared a roadmap on X for caller models, and successful nan comments, said it would beryllium a matter of "weeks/months" earlier GPT-5 was launched. However, successful April, Altman said location had been a alteration of plans, pivoting to a merchandise of o3 and o4-mini and pushing GPT-5 backmost a mates of months, pinch nan exemplary yet launching successful August.
The bigger picture
Ultimately, nan accelerated gait astatine which OpenAI is moving aligns good pinch nan company's eventual extremity of achieving AGI. Although astir estimates still spot AGI acold away, OpenAI's eager extremity does promote them to perpetually innovate, research, and make incremental upgrades. Releasing truthful galore models has fed nan company's upward trajectory of momentum.
Also: Claude wins precocious praise from a Supreme Court justness - is AI's ineligible losing streak over?
The Information recently reported that OpenAI reached $12 cardinal successful revenue, double nan already staggeringly precocious $5.5 cardinal reported successful December 2024 (though nan institution is still not profitable). The study besides noted that OpenAI has astir 700 cardinal progressive users for its ChatGPT products, including endeavor and business customers.
OpenAI has been riding this activity for immoderate clip now, pinch a previous study from The Information showing that, successful nan first 3 months of 2025, nan institution saw a 30% summation successful revenue. This maturation is uncharacteristic of nan larger industry, because group are overwhelmed by the sheer number of offerings on nan market.