OpenAI’s most succesful AI mannequin, GPT-5, could also be coming in August

References to “gpt-5-reasoning-alpha-2025-07-13” have already been noticed on X, with code exhibiting “reasoning_effort: excessive” within the mannequin configuration. These sightings counsel the mannequin has entered last testing phases, with testers getting their fingers on the code and safety consultants doing crimson teaming on the mannequin to check vulnerabilities.

Unifying OpenAI’s mannequin lineup

The brand new mannequin represents OpenAI’s try and simplify its more and more complicated product lineup. As Altman defined in February, GPT-5 might combine options from each the corporate’s typical GPT fashions and its reasoning-focused o-series fashions right into a single system.

“We’re actually excited to not simply make a internet new nice frontier mannequin, we’re additionally going to unify our two collection,” OpenAI’s Head of Developer Expertise Romain Huet mentioned at a latest occasion. “The breakthrough of reasoning within the O-series and the breakthroughs in multi-modality within the GPT-series will likely be unified, and that will likely be GPT-5.”

In keeping with The Info, GPT-5 is anticipated to be higher at coding and extra highly effective total, combining attributes of each conventional fashions and SR fashions akin to o3.

Earlier than GPT-5 arrives, OpenAI nonetheless plans to launch its first open-weights mannequin since GPT-2 in 2019, which implies others with the right {hardware} will be capable of obtain and run the AI mannequin on their very own machines. The Verge describes this mannequin as “just like o3 mini” with reasoning capabilities. Nevertheless, Altman introduced on July 11 that the open mannequin wants extra security testing, saying, “We aren’t but certain how lengthy it can take us.”