OpenAI’s next flagship model may not represent as big a breakthrough as its predecessors. a new report In Information.

Employees who tested the new model, codenamed Orion, reportedly noted less improvement than what they saw in the transition from GPT-3 to GPT-4, although its performance exceeded OpenAI’s existing models.

In other words, the rate of recovery appears to be slowing. It might actually be Orion Negative It can be reliably better than previous models in some areas, such as coding.

In response, OpenAI has assembled a core team to understand how the company can continue to improve its models in the face of dwindling new training data. These new strategies reportedly involve training Orion’s AI models on synthetic data generated by them and developing further models in the post-training process.

OpenAI did not immediately respond to a request for comment. In response to previous reports regarding plans for the flagship model, the company said“We have no plans to release a model codenamed Orion this year.”