CEO Operanai SAM Sam Altman says the company is forced to crawl rollouts from the latest models, Gpt-4.5Due to the opening “from the GPU.”
In a post on xAltman says GPT-4.5, described as a “giant” and “expensive,” will require “tens of thousands of GPUs can get access. Gpt-4. 4. 4 The first time it will come to the customer for the Pro chatter start Thursday, followed by customers next week.
Perhaps it is part of a very large size, GPT-4. 4. 4.5 is precious. Openai charge $ 75 per million token (~ 750,000 words) food to model and $ 150 per million tokens made by the model. That 30x costs input and 15x cost output from opening workhorse Gpt-4o Model.
Gpt 4.5 harn. If this does not have a very large model I will be disappointed pic.twetter.com/1kk5lpn9Gh
– Casper Hansen (@casper_hansen_) February 27, 2025
“We’ve had much and no GPU,” wrote. “We will add tens of GPUs next week and plump to Tier Plus then (…) This does not want to use, but it’s difficult to predict the GPU lack.”
Altman has Formerly says that the lack of computing capacity is the delay of a company’s product delay. Openai hopes for this combat in the coming years Developing your own AI chipsand by Build a number of data dataSee rankings-.
