Anthropic’s latest flagship AI might not have been incredibly costly to train

The new flagship AI model of anthropic, Claud 3.7 Sonnet, spends “several million dollars” for training using less than 10^26 computing power.

According to Warton Professor Ethan Molik, who expressed a clarity from anthropological public relations in an X post on Monday. “I was an ethnographic contact who told me that Sonnet 1.7 would not be considered as a 10^26 flop model and would spend a few million dollars,” He wrote“Though the future models will be much larger.”

TechCrunch reached the ethnicity for confirmation but did not receive any response as time to publish.

Claud 4.7 Spending on Sonnet spent “several million dollars” for training without factoring at the Sonnet, which is a sign that is becoming relatively cheap to publish sophisticated models. Claud 3.5 Sonet’s predecessor, published in the fall of 2024, Similarly is spent on several million dollars for trainingAnthropological chief executive officer Dario Amodai published in a recent article.

These total are quite favorable with the training price tags of the top models of 2023. OpenAI has spent more than $ 100 million to develop its GPT -4 model, Accordingly OpenAI CEO Sam Altman. Meanwhile, Google has spent about $ 200 million for training a study in Stanford ApproximatelyThe

Is said that Amodai expects the future AI models Spent billions of dollarsThe Of course, training expenditures do not capture work like security tests and basic research. Furthermore PeriodComputing costs of running models will probably increase.