OpenAI’s latest AI model, GPT-5, has reportedly struggled to meet expectations, raising concerns over its development costs and performance gains, according to a recent report by The Wall Street Journal. This comes on the heels of a similar analysis by The Information, suggesting that OpenAI is exploring new strategies as GPT-5, code-named Orion, might not represent a significant leap over its predecessor, GPT-4.
Development Hurdles and High Costs
The WSJ highlights that GPT-5 has undergone two major training runs during its 18-month development cycle. These runs, designed to refine the model by processing vast datasets, encountered unexpected delays. The initial run progressed slower than anticipated, signaling that subsequent runs would be even more time-intensive and expensive. Despite improvements, the advancements seen in GPT-5 reportedly fall short of justifying its operational costs.
Efforts to Enhance Model Performance
In an effort to boost GPT-5’s capabilities, OpenAI has taken unique steps beyond relying solely on publicly available data and licensing deals. The company has employed teams to generate original content, such as writing code and solving mathematical problems. Additionally, OpenAI is leveraging synthetic data produced by its other AI systems, including a model referred to as o1.
No Immediate Launch for GPT-5
While OpenAI has not confirmed the release of GPT-5 this year, the company’s efforts to address the challenges highlight the complexities of advancing generative AI technology. OpenAI did not provide a comment on the WSJ’s findings.
