Perhaps due to disappointing results, Altman was of Were written earlier That the GPT -4.5 will be the last of the traditional AI models of Openi, with GPT5 plans to be a dynamic combination of “irrational” LLM and artificial reasoning models like O3.
A Stratosopheric Price and a Tech Dead End
And about this price – it’s a grief. The GPT -4.5 costs $ 75 per million input token and an output token of dollars per million via API, compared to GPT -4O’s 50 2.50 million input token and $ 10 per million output token. (There are pieces of data used by AI models for token processing). This of developers using the Open EI model, these prices make the GPT -4.5 unacceptable for many applications where GPT -4O already performs properly.
On the contrary, Openi’s flagship reasoning model, O1 Pro, costs an output token of $ 15 and $ 60 million per million input tokens -which is specifically less than GPT -4.5, which offers special fake reasoning capabilities. Even more amazing, the O3-MINI model costs only $ 1.10 per million input token and $ 4.40 per million output token, making it cheaper than GPT4O, providing stronger performance on specific tasks.
Openi knows how to reduce return to LLM for some time. As a result, the company worked most of the last year on fake reasoning models such as O1 and O3, which use different in conference time (run -time) approach to improve performance rather than throw every large amount of training data in GPT styling AI models.
Although it looks like bad news for the open in the short term, the competition in the AI market is promoting. Anthropic Claude 3.7 Swant With the alleged more efficient architecture, GPT has performed much better than 4.5. It is worth noting that Claude 3.7 Swants are probably a system of AI models working behind the screen, though Anthropic did not provide details about its architecture.
For now, it seems that GPT -4.5 can be the last part of its kind. A technical dead end for the vision of unintentional learning that has paved the way for new architecture in AI models, such as the O3 diagnostic time reasoning and perhaps even more novels, such as Battle -based models. Only the time will tell how matters end.
GPT -4.5 is now available for Chat GPT Pro users, which is planned for Rollout to Plus and team users next week, followed by enterprise and education users after a week. Developers can access it on salaried levels through various APIs of Openi, though the company is uncertain about its long -term availability.