An article summarized by Quartz:

Meta unveiled a new artificial intelligence model called Muse Wonder on Monday, claiming it can compete with leading models from OpenAI’s ChatGPT and Anthropic’s Claude on tasks such as coding and answering complex financial and legal questions. Meta AI chief Alexandr Wang described the model as a frontier-level system designed to perform well on both personal and professional tasks. However, independent testing has yet to establish whether its real-world performance matches Meta’s claims, as AI benchmark results do not always reflect how models perform in everyday use.

The launch represents a major step in Meta’s effort to strengthen its position in the AI race after its earlier models received a disappointing reception. Internally known as “Watermelon,” Muse Wonder is part of a 14-month overhaul led by CEO Mark Zuckerberg, who has committed hundreds of billions of dollars to AI infrastructure and brought Wang into the company through a reported $14 billion deal involving Scale AI. Unlike competitors that have concentrated heavily on coding tools, Meta says it is prioritizing personal AI agents that can help users complete everyday tasks.

Muse Wonder is currently available to developers through Meta’s Model API and Muse Code, with plans to integrate it into the subscription version of the Muse AI agent app. Launched on September 8th, Muse reportedly reached five million downloads in 22 days, according to Sensor Tower, outpacing the early adoption of ChatGPT, Grok, and Claude. The app can perform tasks such as sending emails, booking travel, filling out forms, and making purchases, with subscription plans priced at $20 and $100 per month. Meta is betting that combining powerful AI models with practical, task-oriented assistants will help it compete more effectively in the rapidly growing AI market.

For the article: https://qz.com/meta-muse-wonder-ai-model-launch-101226