Running any task with the ChatGpt O3 model can be up to 5 times more expensive than the initial estimation

OpenAI collaborated last December with the introduction of the O3 artificial intelligence model with the “Arc-Agi” benchmark to better display its capabilities. Now, after a few months, the results of this benchmark show that the performance of this model is not as significant as expected.

Running each task at ChatGpt O3 costs about $ 30,000

Last week, the Arc Prize Center, which performs the benchmark tests, updated the O3 computational costs. The foundation initially estimated that solving a problem with Arc-Agi tests and using the strongest O3 model, the O3 High, would cost about $ 3,000. But now in its latest research, the center has announced that implementation of each task could cost about $ 30,000.

Openai has not officially released the O3 yet, but the center believes that the new model should be close to the O1-Pro model, the most expensive model of the Openai artificial intelligence. “We think the O1-Prize model is the closest model in terms of cost, because both have similar computing volumes to execute tsors,” said Mike Nop, a co-founder of the Arc Prize.

Of course, given the volume of processing and O3 calculations, Openai is unlikely to be high. Of course, the model is tested in two versions “O3 High” and “O3 Low”, each of which has a different computational capability, as its name implies.

In recent times, there have been rumors about expensive Openai projects for corporate customers. Earlier, The Information website said in a report that Openai may receive up to $ 20,000 a month to use artificial intelligence -based factors.

While the use of artificial intelligence agents is more affordable than hiring human resources, Toby Ord, a researcher at the X platform, pointed out that these models are not necessarily higher than human resources. For example, the O3 High model had to repeat the processes three times to deliver the best possible results in the task.