OpenAI to launch GPT-5 in August: Changes, improved performance and all that’s expected

ChatGPT

OpenAI is preparing to unveil its next-generation large language model, GPT-5, in early August. The company’s CEO, Sam Altman, recently confirmed that GPT-5 is “coming soon,” offering hints about its expanded capabilities and internal testing. The update is expected to include substantial improvements over GPT-4, with new variants designed for different usage scenarios.

While GPT-5 was originally expected in May 2025, internal testing extended the timeline. Now, multiple sources suggest that the official rollout is planned for August. Ahead of the launch, GPT-5 has reportedly been used privately, including by Altman himself, who shared a recent interaction where the model outperformed his expectations during a podcast appearance.

Unlike previous updates, GPT-5 is expected to incorporate “o3” reasoning capabilities directly within the model. This suggests a more unified approach to AI development at OpenAI, reducing the need to split features across separate models. The shift is also likely to improve performance in tasks that require logical deduction, planning, and problem-solving.

In addition to the standard GPT-5, OpenAI is expected to release smaller versions of the model—tentatively named “mini” and “nano.” These versions are aimed at developers who need efficient models for on-device or edge computing, and they will be made available via OpenAI’s API platform. This follows the trend of offering flexible model sizes for a variety of applications.

During a recent appearance on This Past Weekend, Altman recounted using GPT-5 to solve a technical problem he didn’t understand. The model delivered a perfect response, prompting Altman to reflect on the growing gap between human and AI capabilities. He described it as a “here it is” moment, highlighting the model’s intuitive understanding and reasoning.