Meta Platforms has shipped its long-delayed flagship AI model, now called Muse Spark, ending months of speculation after the release slipped three separate times. The model was known internally as Avocado during development before Meta settled on its public name.

The launch comes as Meta pushes further into AI infrastructure beyond its core social apps. Earlier this year, on July 1, the company announced Meta Compute, a new cloud business built to sell excess AI computing capacity to outside customers. Meta is reportedly in early talks to rent computing power to Anthropic, a notable move given that Anthropic is a direct competitor to Meta’s own AI research efforts.

Meta has also expanded its hardware partnerships to support the buildout. The company extended its relationship with Nvidia into a multi-year, multi-gigawatt chip supply deal, a scale of commitment that reflects how much computing capacity Meta believes it will need to support both its consumer AI products and its new cloud ambitions.

The market reaction to Meta’s AI strategy has been largely positive. Meta’s share price climbed above $600 for the first time following the recent wave of announcements, even as the broader run of infrastructure news rattled some chip stocks, which fell on concerns that hyperscalers renting out spare capacity could eventually mean less direct chip demand from those same companies.

Muse Spark’s delayed arrival highlights the pressure Meta has faced to prove its massive AI spending is translating into shippable products, after months in which rivals such as OpenAI, Google and Anthropic released a steady stream of frontier model updates. Meta has not detailed Muse Spark’s full technical specifications or independent benchmark results, though the company has signaled it will provide additional details in the coming weeks as it rolls the model out more broadly across its products.

Meta has not said when Muse Spark will reach general availability across its full lineup of apps, though the company signaled the rollout would happen in stages rather than all at once.