Meta Launches Muse Spark, Its First AI Model Under Alexandr Wang's Superintelligence Labs
What Happened
Meta released Muse Spark, a multimodal model accepting voice, text, and image inputs, marking the first product from its new Superintelligence Labs division led by former Scale AI CEO Alexandr Wang. The proprietary model powers the Meta AI app and website now, with plans to roll across Facebook, Instagram, and WhatsApp. Meta said it could spend up to $72 billion on AI infrastructure this year.
My Take
The headline is the model. The real story is that Meta just abandoned the open-source playbook that defined Llama. Muse Spark is proprietary. That is a significant strategic reversal from the company that spent years positioning itself as the open-weights alternative to OpenAI and Anthropic. For builders, this matters because Meta's open models were the escape valve — the thing you could self-host, fine-tune, and build on without platform risk. If Meta is pulling back from open, the competitive dynamics of the model layer just shifted. Fewer credible open alternatives means more leverage for the closed providers. And $72 billion in capex tells you exactly how Meta views the stakes. The question for product engineers is not whether Muse Spark is good. The question is whether the era of viable open-weight frontier models is ending, and what that means for your architecture decisions.
Read Original Source