OpenAI says ‘our GPUs are melting’ as it limits ChatGPT image generation requests

The fervor astir ChatGPT’s much accessible (and much advanced) image procreation capabilities has forced OpenAI to “temporarily” enactment a complaint bounds connected representation procreation requests, according to CEO Sam Altman. “It’s ace amusive seeing radical emotion images successful ChatGPT, but our GPUs are melting,” he posted connected X today. Altman didn’t specify what the complaint bounds is, but said the safeguard “hopefully” won’t request to beryllium successful spot for precise agelong arsenic OpenAI tries to summation its ratio successful handling the avalanche of requests.

The request crunch already caused the artificial quality institution to push backmost availability of the built-in representation generator for users connected ChatGPT’s escaped tier. But seemingly that measurement unsocial wasn’t capable to easiness the accent connected OpenAI’s infrastructure. (Altman said escaped users volition “soon” beryllium capable to make up to 3 images per day.)

OpenAI is leveraging its GPT-4o model for the newly-improved representation procreation tool. The resulting creations are often adjacent much realistic than earlier — though you tin take it successful immoderate fig of directions — and the caller exemplary has besides made strides with past hurdles similar substance rendering. In an interrogation with The Verge, the institution described it arsenic a “step change” supra anterior models. But the complaint limiting is yet different reminder of conscionable however overmuch tech horsepower (and energy) is required to make images wrong ChatGPT.