Meta launches Muse Wonder model to advance its AI assistant

what happened

The Wall Street Journal reports that Meta Platforms released a new artificial-intelligence model on Monday called Muse Wonder. Meta says the system moves it to the frontier of AI, an objective set after the company established Meta Superintelligence Labs last year.

A key test will be whether the model can improve retention and speed adoption of Muse, Meta’s personal-assistant agent launched last month. SensorTower data shows that Muse has reached nearly nine million downloads worldwide, although download growth has slowed and its 14-day retention trails ChatGPT.

Alexandr Wang, Meta’s AI chief, said the model outperformed OpenAI’s Astra and Anthropic’s Fable on several tests. It will soon power Muse for subscribers on both the $20-a-month and $100-a-month plans. According to Meta, the system offers improved memory and focus, along with greater expertise in health, finance and legal matters.

Meta revamped its AI program in spring 2025 after the disappointing release of Llama 4. Chief Executive Mark Zuckerberg brought Wang from Scale AI to lead Meta Superintelligence Labs, pursuing what he calls “personal superintelligence” for the company’s billions of users. Zuckerberg described Muse as the “centerpiece” of that effort last month. OpenAI’s Dots, Google’s Gemini Spark and an agent from startup Instinct are also competing in the market.

why it matters

The launch links Meta’s model-building efforts directly to Muse’s commercial performance. Slowing downloads and weaker 14-day retention than ChatGPT make user engagement an important measure, while rival agents from OpenAI, Google and Instinct increase competitive pressure on Meta.

instruments