Openai officially introduced its new model of artificial intelligence GPT-4.5. However, only PRO subscribers, which pay $ 200 per month, have yet been accessed. A full -fledged release is delayed due to lack of capacities for machine training.
Openai CEO, Sam Altman, explained in his X (Twitter), the account is that launch is gradual due to a deficiency of graphic processors. According to him, the company tries to solve the problem promptly, and next week tens of thousands of new GPU will be added, which will allow to expand access to a model for users with Plus subscription (worth 20 dollars a month).
Openai is currently depends on NVIDIA products, including their advanced Blackwell graphic processors. However, the demand for these chips is so high that they were sold until October 2025. To reduce this addiction, Openai has started developing its own Shi chips in partnership with Broadcom. However, their production will take several years.
An important aspect is the high cost of processing data of the new model. GPT-4.5 has proved to be much more expensive than the previous versions: its processing costs $ 75 per million input tokens and $ 150 per million weekends. For comparison, similar operations in GPT-4O cost only $ 2.50 and $ 10, respectively.
Sam Altman acknowledged that GPT-4.5 is not a revolutionary model in the field of deep reasons and would not set new records. At the same time, he assured that the model has a "special magic", which he had previously felt in any other Openai development.
The company will be able to make GPT-4.5 accessible to the wider range of users in the coming months.
