GPT-5 Reportedly Underperforming: OpenAI's New Model Faces Scrutiny

GPT-5 Development Faces Delays and Cost Concerns

Recent reporting by The Wall Street Journal indicates that OpenAI’s development of its next-generation model, GPT-5, is experiencing delays. The current results are not yet demonstrating a return commensurate with the substantial financial investment.

Confirmation of Strategic Shifts

These findings align with a prior report from The Information, which highlighted OpenAI’s exploration of alternative strategies. The suggestion was that GPT-5 may not deliver the significant advancements seen in previous iterations.

Details on the Orion Project

The Wall Street Journal’s coverage provides further insight into the 18-month development process of GPT-5, internally referred to as “Orion.” Progress has been slower than initially anticipated.

Training Run Challenges

OpenAI has reportedly finished at least two extensive training runs. These runs are designed to enhance model performance through exposure to massive datasets.

The first training run proved to be slower than projected, indicating that subsequent, larger-scale runs would be both time-intensive and expensive to execute.

Performance vs. Cost

While GPT-5 is reportedly showing improvements over its predecessors, its current capabilities haven’t reached a level that justifies the ongoing operational costs.

Innovative Data Acquisition Methods

Beyond utilizing publicly accessible data and licensing agreements, OpenAI is employing novel data acquisition techniques.

• The company has directly employed individuals to generate new data.
• These individuals are tasked with writing code and solving mathematical problems.

Furthermore, OpenAI is leveraging synthetic data produced by its own model, designated as o1.

Response to Inquiries

A request for comment directed to OpenAI has not yet received a response. The organization had previously announced that a model bearing the “Orion” codename would not be released during the current year.