Meta has introduced Muse Spark, a new multimodal AI model developed by Meta Superintelligence Labs. The model can process both text and images, and Meta is positioning it as a core engine built for its own products rather than following the earlier Llama playbook centered on broader openness.
In an announcement published on April 8, Meta said the launch comes under the leadership of Alexandr Wang. It is the first public result from the company’s AI reorganization after spending billions of dollars to rebuild teams and capacity. The release puts Meta back into direct competition with major AI developers.
Muse Spark debuts first on Meta AI in the US
Muse Spark has already gone live in the United States through the Meta AI app and its dedicated website. Meta said the model will later be integrated across WhatsApp, Instagram, Facebook, and Messenger, extending its reach through apps that serve close to 4 billion monthly active users.
The company also plans to make Muse Spark the core model behind its smart glasses. That would give the wearable lineup stronger visual and voice interaction features. The launch is not limited to software. It also ties Meta’s AI push more closely to its hardware strategy.
No public open weights, only a private API preview
Meta has not released Muse Spark as open weights for the public. Instead, access is currently limited to a Private API Preview offered to selected partners. That marks a clear shift from the Llama series, which became known for its open distribution strategy.
The change has sparked discussion across developer circles and platforms such as Hacker News. Much of that discussion focuses on whether Meta is rethinking how far it wants to go with open access as it pursues superintelligence, and whether it now wants stronger models to serve revenue goals and product lock-in first.
Meta signals a product-first AI strategy
Meta said it has spent the past nine months committing major compute and resources to develop Muse Spark. With the model now public, the company is making its direction easier to read: less emphasis on open community positioning, more focus on deploying AI inside its own apps, services, and devices.
Meta also said Muse Spark is only the first step in the broader Meta Superintelligence Labs model lineup. No wider release plan has been announced so far, but the company has made one point clear: this model is being built to power the next round of AI features across Meta’s platforms and wearables.