On Wednesday, OpenAI CEO Sam Altman announced via his X account that GPT-4.5 had been released for Plus subscribers. This comes just a week after the model was made available to a select group of users and Pro subscribers for research preview testing. After the announcement, Altman cryptically said, “I love this model,” without further details.
GPT-4.5, the successor to GPT-4 launched last year, is expected to be OpenAI’s final non-inference model.
Starting with GPT-5, OpenAI plans to merge inference AI models with general AI models. The company clarified that GPT-4.5 does not replace the current base model, GPT-4.
However, GPT-4.5 has been repeatedly criticized for its poor cost-effectiveness. Industry experts predict that, to boost GPU availability, OpenAI will likely discontinue support for the 4.5 model once GPT-5 is released in May.
When announcing GPT-4.5’s release, Altman commented on the downside of CPU resource shortages, describing GPT-4.5 as a “giant” and “expensive” model. He stated that while they prefer not to operate this way, accurately predicting GPU shortages and surging resource demands is challenging.
Despite these challenges, Altman expressed enthusiasm, stating that he had been surprised multiple times by the advice from this new model. He added, “It represents a different kind of intelligence, with a magical quality I’ve never experienced before. I hope many people can benefit from it.”