Meta Says New AI Model Is State-of-the-Art, Will Make Muse Agent Smarter

Meta Platforms META 4.24%increase; green up pointing triangle released a new artificial intelligence model Monday that it says vaults the company to the AI frontier, a goal it established last year when it created Meta Superintelligence Labs.

One important test of the model, called Muse Wonder, will be how much it can improve retention and accelerate adoption of Muse, the company’s new personal-assistant agent introduced last month.

Muse has amassed nearly nine million global downloads since its release. But downloads have started to slow and its retention lags that of ChatGPT after 14 days, Sensor Tower data shows.

“Some people have a pretty clear sense of what they want to use the product for or how it can improve their lives,” Alexandr Wang, Meta’s AI chief, said in an interview. “Other folks who maybe are less familiar with AI or agents may not understand that, and for them, I think it’s more of a hurdle to get on board.”

“We think it is competitive as a frontier model with the very best models that are out there,” Wang said. “But in addition to that, it also is specifically the very best model for personal and professional agents and will be the best model to power an experience like Muse.”

Meta overhauled its AI efforts in spring 2025 after the disappointing launch of its then-latest model, Llama 4. Meta Chief Executive Mark Zuckerberg hired Wang from Scale AI to lead the effort at newly created Meta Superintelligence Labs with a vision of providing so-called “personal superintelligence” to its billions of users.

That goal has emerged in the form of Muse, which Zuckerberg said at a conference last month is the “centerpiece” of what the company is building.

OpenAI released Dots, a competing AI agent, at its annual developer conference nearly two weeks ago. Google offers an agent called Gemini Spark and startup Instinct has also launched its own buzzy agent that has emerged as the main competitor to Muse so far.