Anththropic’s New Flagship Ai Model, Claude 3.7 Sonnet, Cost “Tenses a few of millions of dollars” Training using less than 10 ^ Computing chops.
That’s according to Shatton Professor in Ethan Mollick, sent by X on Monday to refer clarification that you had received from anthropic’s Pr. “I have been touched by anthropic who told me Sonnet 3.7 will not be viewed as a 10 ^ 26 Flop and costs several millions of dollars,” writes,
Techcrunch has been reached with anthropic for confirmation but has not received a response such as the time of publication.
Taking Claude 3.7 Sonnet is actually expensive “for several millions of dollars” training, not to save the costs related to, it is a sign of how cheaper is the back of humanity. Claude 3.5 Cloud 3.5, has been issued at WIL 2024, like a number of decades for millions of dollars to train, Anthropic CEO Dario Amode are presented in the latest story.
Those parts are well compared to prices for training for 2023 models. Developing its GPT-4 model, Openaai spend more than 100 million, according to Opelai Ceo Altman. At that time, Google dismisses about $ 200 million to train its model ultra, limited Stanford Survey.
That is said, Amodei expects future models to call billions of dollars. Certainly, the training costs do not disclose a function such as safety and basic research. In addition, as ai industry includes the “Reasoning” models they apply to strangers of extended periods, running computer costs will continue to rise.
Source link
