OpenAI’s Latest Model, GPT-5: Features, Feedback, and Future Updates

OpenAI recently unveiled its latest language model, GPT-5, in a Reddit AMA session with CEO Sam Altman and key team members. The new model introduces a feature that determines the best model to use based on the prompt, aiming to provide either quick responses or more thoughtful answers. However, some users expressed that GPT-5 appeared less effective than its predecessor, GPT-4o.

Altman clarified that initial issues were due to a malfunctioning router during the rollout, which made GPT-5 seem less capable. He assured users that improvements are underway, including adjustments to the decision-making process for model selection and increased transparency about which model is responding to queries.

Amidst user requests, Altman mentioned that they are considering allowing Plus subscribers to access GPT-4o while gathering more data on the trade-offs involved. Furthermore, OpenAI plans to double the rate limits for Plus users, facilitating more extensive use of GPT-5 without prompt limitations.

A notable incident during the model’s presentation was a chart error that misrepresented benchmark scores. Although Altman did not directly address this during the AMA, he acknowledged the mistake earlier and promised corrections.

Overall, OpenAI remains committed to refining GPT-5 based on user feedback and ensuring the model’s stability and reliability. As the rollout progresses, users can expect ongoing enhancements and adjustments to optimize their experience with GPT-5.