OpenAI Enhances GPT-4o and GPT-4-mini-high Models for ChatGPT Plus Users

OpenAI has increased the hourly rate limits for its GPT-4o and GPT-4-mini-high models, available to ChatGPT Plus subscribers. This modification aims to alleviate restrictions for users who require enhanced capabilities. CEO Sam Altman verified the update on X, indicating that this change was made in direct response to user opinions.

The enhancement allows paying subscribers to send and receive a significantly higher number of messages per hour, thus bettering the service for those engaged in high-frequency tasks, whether it be coding, researching, or creating content.

Part of Altman’s statement read, “This is part of our continuous endeavour to enhance the ChatGPT experience based on your feedback.” He recognised that while the company is responsive to user input, striking a balance between scaling services and managing technical limitations is a challenging task.

Infrastructure Challenges at OpenAI

OpenAI has highlighted ongoing challenges related to infrastructure, particularly concerning the availability of GPUs essential for operating its large language models at scale. Altman remarked that the company is continually facing “hard tradeoffs” between enhancing model accessibility, preserving performance, and innovating new functionalities.

He further commented, “Demand is high and GPUs are still scarce,” noting that OpenAI is making strides to incorporate “tens of thousands of GPUs” to alleviate the situation.

Transition to GPT-4o Model

In conjunction with the modifications to rate limits, OpenAI is set to phase out the original GPT-4 model in ChatGPT. Effective April 30, GPT-4o will completely replace GPT-4 as the default model for all ChatGPT users. This transition is expected to streamline the user experience while also consolidating demand onto a singular, infrastructure-intensive model.

The company perceives this shift as a component of its broader strategy to optimise performance while effectively managing the substantial resource requirements involved in operating multimodal AI systems.