Thought Leadership
OpenAI is about to release its latest language model GPT-5, which is expected to come onto the market in August. This was reported by The Verge, citing informed sources from the company.
The launch was originally planned for the end of May, but had to be postponed due to extensive additional tests and technical optimizations.
Altman demonstrates new capabilities
CEO Sam Altman gave first insights into the performance of the upcoming system on the social media platform X. In a recent podcast appearance on “This Past Weekend”, Altman had GPT-5 answer a complex question live that he was unable to solve himself. “I put it into the model – that’s GPT-5 – and it answered perfectly,” explained the OpenAI boss. The demonstration gave him a strange feeling, as he felt “useless in the face of AI”.
According to The Verge, OpenAI is working on a new model generation, which is currently called GPT-5 and is to be released in various sizes. A standard, a “mini” and a “nano” version are planned, all of which will be offered via OpenAI’s API.
The main model will feature both traditional speech processing and simulated reasoning capabilities and will be available both in ChatGPT and via the API. The smaller Nano version, on the other hand, will only be accessible via the API, not directly in the ChatGPT interface.
Model consolidation as a strategic step
With GPT-5, OpenAI is pursuing a new product strategy: instead of developing separate model series, the company is integrating its o3 reasoning technology directly into the main version. “It is a system that combines many of our technologies,” Altman described the new flagship model at the beginning of the year.
This consolidation should not only improve the user experience, but also advance OpenAI’s long-term goal of achieving Artificial General Intelligence (AGI). AGI certification would have far-reaching consequences for the Microsoft partnership: if this milestone is reached, the tech group would lose its rights to OpenAI’s revenue streams and future AI models.
According to The Information, GPT-5 should be able to code better and be more powerful overall, as it combines the features of both traditional models and SR models such as o3.
However, Altman warned against exaggerated expectations at the time of launch. The “gold level of capabilities” will only be reached “many months” after release.
Open source preview before main release
As an appetizer, OpenAI is planning to release an open-weight model before the GPT-5 launch – the first since GPT-2 from 2019. The system, described as “similar to o3 mini”, is expected to be available in July via Azure, Hugging Face and other cloud providers and will also have reasoning capabilities.
