Following the success of ChatGPT, OpenAI has introduced many other tools for users to choose from. Among them is GPT-5 - the next potential version in the GPT language model series.

Codenamed Orion, GPT-5 was scheduled to launch in mid-2024, but the project has been delayed by a number of issues, mostly related to technical issues, costs, and training data. Leaked information shows that GPT-5 training has not met initial expectations, with two main training rounds yielding poor results.

The reason given was that the quality and variety of the training data did not meet the requirements for this new version. The lengthy testing process made the development process difficult and also increased the cost. It is known that a single GPT-5 training run costs about half a billion dollars in computing resources.

OpenAI faced resource and financial constraints when developing its new language model. Collecting data from the internet was not enough, so the company created custom data files from scratch, which improved the quality and diversity of the data.

OpenAI hired experts from many fields to not only write code but also solve problems and explain complex ideas. This makes for rich data but is also expensive and time-consuming. That's why GPT-5 is still delayed.