OpenAI Struggles with GPT-5: Delays and High Costs Slow Development | ChatGPT app | OpenAI stock | OpenAI free | Turtles AI

 OpenAI’s development of GPT-5, codenamed “Orion,” is behind schedule, with performance not justifying the costs incurred. Technical problems and innovative strategies characterize the project.

Key points:

• Slow progress: GPT-5 does not yet represent a significant technological leap from previous models.
• High cost: Training sessions are slow and resource-intensive.
• Synthetic data: OpenAI is using internal models and dedicated teams to generate new data.
• Alternative strategies: The company is adopting unconventional solutions to address technical challenges.

According to a Wall Street Journal report, OpenAI is encountering significant obstacles in developing its GPT-5 model, known internally as “Orion.” The creation phase, which began 18 months ago, appears to be proceeding more slowly than expected, highlighting both technical and economic difficulties. Although GPT-5 may have improvements over its predecessors, its performance is not yet innovative enough to justify the massive investment required for training and maintaining the model.

During development, OpenAI completed at least two major training cycles, processes that are fundamental to optimizing an AI’s capabilities by processing huge datasets. However, one of these cycles turned out to be slower than expected, a troubling sign for the company, which faces the prospect of further delays and rising costs. The difficulties are not limited to training: the model, while improved, still fails to clearly outperform previous models such as GPT-4, raising questions about the sustainability of these efforts.

In addition to using existing public data and licenses, OpenAI chose to take a more direct and experimental approach to enrich GPT-5 training. It hired experts to create custom data, producing code and solving complex mathematical problems, and leveraged synthetic data generated by another of its models, identified as “o1.” This method aims to compensate for gaps in traditional datasets, but it is not without its limitations and risks.

Despite the attention generated by the project, OpenAI has stated in the past that it does not intend to release a model codenamed “Orion” by the end of the year. At present, the company has provided no further official details in response to media inquiries.

Regardless of the outcome, the development path of GPT-5 reflects the growing challenges in advancing AI.