Meta Superintelligence Labs just dropped its first model since Mark Zuckerberg poured billions into restructuring the company’s AI division. The model is called Muse Spark, and it’s already live in the Meta AI app and website in the US, as reported by The Verge AI.
This isn’t just another model refresh. It’s the start of an entirely new series called Muse, separate from Meta’s Llama lineup, which stumbled badly with the delayed Llama 4 release in 2025.
What Muse Spark Actually Does
Meta describes the model as “purpose-built for Meta’s products,” drawing an obvious parallel to how Google Gemini threads through Google’s ecosystem. Here’s what’s shipping:
- Multimodal input supporting both text and images
- Two reasoning modes: a fast “Instant” mode and a slower “Thinking” mode for complex queries (similar to Microsoft’s Think Deeper)
- Multi-agent architecture that runs AI sub-agents to handle queries faster
- Health-related capabilities including image and chart interpretation
- Private API preview for select partners
The rollout to WhatsApp, Instagram, Facebook, Messenger, and Meta’s smart glasses is coming “in the coming weeks,” along with international expansion.
Why This Matters
Meta is essentially admitting that Llama alone wasn’t enough. After Llama 4’s rocky launch forced Zuckerberg to overhaul the entire AI program, Muse Spark represents the company’s second serious attempt at building competitive AI infrastructure.
The multimodal angle is particularly interesting for Meta’s Ray-Ban smart glasses play. If the company is serious about AI-powered glasses being the next computing platform, it needs a model that can see and reason about the real world. Muse Spark is designed to do exactly that.
The Health AI Angle
Meta is wading into health-focused AI, a space that’s both lucrative and risky. The company highlighted Muse Spark’s ability to “navigate health questions with more detailed responses, including some questions involving images and charts.” It’s clearly eyeing OpenAI’s ChatGPT Health and Anthropic’s Claude for Healthcare, both launched in January.
Health AI chatbots remain controversial. They handle sensitive personal data and can spread misinformation. Meta’s demo showed calorie estimation for meals, a feature that’s popular but notoriously unreliable across AI platforms.
What Comes Next
Meta called Muse Spark an “early data point” on its Muse trajectory. Larger models are in development, and the company says it hopes to open-source future versions. There’s also a plan to integrate recommendations and content people share across Instagram, Facebook, and Threads into the model’s responses.
What stands out here is the speed of the pivot. Zuckerberg didn’t just patch Llama. He built a new lab, a new model series, and a new distribution strategy. Whether Muse Spark can actually compete with OpenAI, Google, and Anthropic remains the open question, but Meta’s 3.5 billion daily active users across its apps give it a distribution advantage that nobody else can match.
Full details are available at the original report from The Verge AI.