@OpenAI may have reallocated resources from GPT-4o and directed them towards their new model.

GPT-4o appears to show similar issues that plagued the later versions of GPT-4, such as slower predicting, laziness, network errors, and increased hallucinations, likely due to insufficient resources.