GPT-5 is Coming: OpenAI’s Next-Gen Model Set for August Launch with “O3” Reasoning & Mini Variants

OpenAI is preparing to unveil its next-generation language model, GPT-5, with the official release expected as early as August. Earlier this year, reports emerged that Microsoft engineers had begun provisioning server infrastructure in anticipation of the new model, which was initially slated for a May launch. However, extended testing led to a postponement.

OpenAI CEO Sam Altman recently confirmed on X that “we’ll be releasing GPT-5 soon,” and in a podcast with Theo Von, he reflected on its capabilities. He recounted how he presented the model with a problem he himself did not understand, only to be astonished by how swiftly and precisely GPT-5 resolved it: “That was the ‘aha’ moment,” he admitted. “I felt utterly useless beside it.”

Even before Altman’s public acknowledgment, whispers of GPT-5 had already begun circulating. Users shared screenshots and benchmarks suggesting its existence, further fueling speculation about an imminent release. According to sources, OpenAI plans to launch not only the primary model in early August but also lightweight variants — mini and nano — both of which will be accessible via API.

Altman previously described GPT-5 as “a system that integrates many of our technologies.” It will encompass the reasoning capabilities previously associated with the o3 family, which will no longer be released as a standalone model. This marks a significant step toward architectural unification at OpenAI — a move aimed at constructing a more robust system potentially capable of laying the groundwork for Artificial General Intelligence (AGI).

Why this matters: achieving AGI-level capabilities could fundamentally alter the dynamics of OpenAI’s partnership with Microsoft. As reported by The Information, the formal recognition of AGI would terminate Microsoft’s rights to future profits and model access. However, Altman has clarified that GPT-5 will not reach this “gold standard” of capability for “many months” post-launch.

Additionally, the merging of GPT and o-series lines will spare users from having to select different models for different tasks. The full-featured GPT-5 — complete with advanced reasoning — will be available through both ChatGPT and the API. The mini variant will also appear on both platforms, while nano will remain exclusive to the API.

Although the official launch is anticipated in early August, OpenAI is known to adjust its release timeline based on model readiness, server infrastructure capacity, and competitive developments. Earlier in July, the company postponed the release of its new open-weight language model to allow for further evaluations and risk assessment.

Nevertheless, the latest reports suggest that this open-weight model is still on track to be released by the end of July. It is described as akin to o3 mini — equipped with contextual reasoning, inference, and problem-solving abilities, rather than mere pattern recognition. This will mark OpenAI’s first openly accessible model since GPT-2 in 2019. For researchers and developers, this means the ability to download the model weights and run it on their own infrastructure. Distribution is expected via Azure, Hugging Face, and other major cloud platforms.