The GPT-4o model has taken image generation in ChatGPT to a new level, but its success has forced OpenAI to impose restrictions due to the overload on its systems.
The GPT-4o model, introduced by OpenAI on March 25, 2025, has revolutionized the artificial intelligence image creation experience within ChatGPT, thanks to substantial improvements in realism, visual consistency, and text rendering. However, the massive enthusiasm for this new functionality has created an unexpected side effect: the saturation of the company’s technical infrastructure.
In light of this situation, OpenAI has announced the implementation of a temporary usage limit on the free version of the service. Starting in the coming days, users will be able to generate a maximum of three images per day for free. This measure aims to alleviate the load on the servers while a long-term solution is developed.
“Our GPUs are melting”
This was acknowledged by Sam Altman, CEO of OpenAI, in a message posted on social media platform X, where he explained that demand has far exceeded expectations:
“It’s very exciting to see how much people enjoy the images in ChatGPT… but our GPUs are melting. We’re going to introduce temporary limits while we work on improving system efficiency.”
The new image generation system has dramatically increased the number of requests from around the world. In forums like Reddit, users continually share examples of images created with GPT-4o, ranging from scenes in the style of Studio Ghibli to concept designs and hyper-realistic compositions.
Advanced technology available to everyone (for now)
GPT-4o marks a significant improvement over previous versions: it can interpret natural language descriptions more accurately, generate readable text within images, and produce results that approach a professional level. This advancement has made the tool one of the most used in the OpenAI ecosystem in just a few days.
The system also integrates directly into ChatGPT in its web and mobile versions, making access to image generation as simple as writing a sentence.
Some glitches and promise of improvements
OpenAI has also acknowledged that, amid the surge in requests, there have been technical errors that prevented some users from generating images. The company assures that it is working to resolve these issues as quickly as possible.
For now, unrestricted access remains available for ChatGPT Plus and Enterprise subscribers, while users on the free tier will need to adapt to the new daily limits.
The challenge of scaling generative AI
This episode brings to light the challenges of scaling generative artificial intelligence for the general public. Although OpenAI has proven to be at the forefront of innovation, the exponential growth in the use of its services poses increasing demands in terms of computing power, energy efficiency, and responsiveness.
With an increasingly active and creative user community, all signs indicate that such measures could be repeated if there is not consistent investment in high-performance infrastructure. Meanwhile, the GPT-4o model continues to demonstrate that generative AI is reaching a level of maturity that redefines the limits of what is possible.
Source: Artificial Intelligence News