The news: On Wednesday, Meta launched Muse Spark, the first in a new series of large language models (LLMs) from its Superintelligence Labs (MSL) and what the Big Tech player calls its most powerful model to date.

Muse Spark now underpins the Meta AI app and website and will roll out to WhatsApp, Instagram, Facebook, and Messenger—as well as Meta’s AI glasses—in the coming weeks.

The model can take video, image, and text inputs, though it can only output text. Its features include:

• Shopping mode: Draws context from creators and communities that users already follow to provide styling or product recommendations.
• Visual coding: Creates custom websites and mini-games from user prompts.
• Health guidance: Offers responses on common health questions and concerns, such as calorie information based on a photo of food.

Muse Spark competes with products from Google and OpenAI, though there are gaps in areas like coding workflows and complex multistep agentic tasks compared with models from outside labs, per the company. It’s a step toward CEO Mark Zuckerberg’s goal to create an AI superintelligence that can act as the ultimate personal assistant.

Zooming out: MSL—a specialized AI division at Meta that was established to develop artificial general intelligence (AGI)—has seen several upheavals since launching in June 2025. That includes an August reorganization that split it into four units, focused on research, superintelligence, products, and infrastructure.

Muse Spark could put Meta back in the AI race after its previous model, Llama 4, received a lukewarm reaction, per Fortune.

Why it matters: Muse Spark shows that Meta is making progress on its AI endeavors, despite challenges in the AI race—such as model delays and MSL layoffs that scaled down focus on areas like the metaverse and VR as a whole. It plans to invest $600 billion to support AI growth by 2028.

The company said models are scaling “predictably,” with Muse Spark serving as an early data point on its trajectory.

• Muse Spark’s broad, generalist capabilities—which span planning, visual interpretation, and everyday decision assistance—highlight Meta’s focus on embedding AI across all consumer touchpoints.
• Rather than competing purely on technical benchmarks, Meta is positioning its models around utility and distribution, using its app ecosystem to drive adoption.

While Meta plans to release a version under an open-source license, it’s currently an in-house tool that’s otherwise only available to select partners.

Implications for brands: Personalization is one of consumers’ top-requested features in AI tools, and Muse Spark’s utilitarian and assistant-first design could accelerate expectations for always-on, highly tailored brand interactions across Meta’s platforms.

However, as long as Meta keeps the model largely within its own walled garden, brands may face more limited visibility and control over how their products are surfaced on Meta apps, forcing a greater reliance on Meta’s AI-driven discovery and recommendation systems.

This content is part of EMARKETER’s subscription Briefings, where we pair daily updates with data and analysis from forecasts and research reports. Our Briefings prepare you to start your day informed, to provide critical insights in an important meeting, and to understand the context of what’s happening in your industry. Non-clients can click here to get a demo of our full platform and coverage.