Meta Announces Its ‘Strongest’ AI Model Yet, Emphasizes ‘Aggressive And Attractive’ Pricing

Meta released an update to its Muse Spark artificial intelligence model, with top official Alexandr Wang calling its pricing strategy as “aggressive and attractive.”

The model in question is Muse Spark 1.1. The company is making its API available through a developer portal as part of a public preview, CNBC noted. There, users can sign up and get steps for integration. For now, the company is limiting to its own properties and not making it available on third-party platforms.

“This is going to be served on top of the computer infrastructure that we’ve built,” said Wang. Elsewhere, he said API accounts will start with $20 in free credits, then charging $1.25 per million token inputs and $4.25 per million tokens of output.

The company is also preparing to launch a cloud computing business aimed at selling excess artificial intelligence computing capacity as it seeks to capitalize on its massive investment on AI. The company has projected between $125 billion and $145 billion in capital expenditures this year, one of the largest technology investment programs in corporate history.

Bloomberg News detailed last week that Meta is developing a cloud business that would allow outside customers to access the company’s AI infrastructure and computing power, clarifying that the plan remains under development and could still change.

The strategy puts Meta in direct competition with established cloud giants such as Amazon Web Services, Microsoft Azure and Google Cloud, while also challenging specialized AI infrastructure providers including CoreWeave and Nebius.

Meta’s cloud platform could allow developers to access AI models running on the company’s infrastructure while paying for the computing power required to operate them. The company is also reportedly considering selling raw AI compute capacity, similar to the services offered by emerging “neocloud” providers.

The move could also help Meta diversify beyond its advertising business, which still generates the overwhelming majority of its revenue.

Meta’s potential entry into cloud computing has not come entirely out of nowhere. During the company’s annual shareholder meeting in May, Zuckerberg acknowledged that launching a cloud business was “definitely on the table” if Meta eventually built more computing capacity than it needed internally.

Meta is also training a more powerful AI model, code-named Watermelon. It has not disclosed when it will be released.