It has been over a year since OpenAI launched GPT 4, and while we did receive GPT 4 Turbo later on, it was less of a successor and more of an upgrade over the existing GPT 4. This is why many are expecting to see a true successor this year, but it is not going to be GPT 5.
Due to the sheer amount of computing power required for such a model, analysts believe that GPT 5 will not happen this year.
In the words of Dan Hendrycks, head of the Center for AI Safety, every time OpenAI upgrades its GPT LLM, it demands ten times more computing power. So, skipping GPT-4.5 and moving straight to GPT-5 would mean needing about a hundred times more computing power than for GPT-4. To put it in perspective, that’s like running approximately 1 million H100 chips continuously for three months. The analyst doubts that OpenAI has a 1 million GPU server already.
Backing this claim is the insight shared by Anthropic’s CEO, Dario Amodei. He recently highlighted that training a cutting-edge LLM currently carries a price tag of approximately $1 billion. However, this expense is predicted to surge to between $5 billion and $10 billion by 2025/26. Significantly, the $1 billion training cost mirrors the tenfold increase in computational resources, which is a reasonable assumption for GPT-4.5.
GPT 4.5 is more likely to be released this year since it has already appeared in leaks coming from the OpenAI blog. Last month, search engine results showed a webpage from the OpenAI blog detailing GPT 4.5, which is poised to be faster, more accurate, and more scalable.
It will have double the context window of GPT 4 Turbo at 256,000 input tokens and an updated knowledge cut-off date of June 2024, which is why it is believed the GPT 4.5 is coming in June.
