On Wednesday, Meta unveiled a new AI model named Muse Spark, representing its “initial move” towards a significant “revamp of [its] AI initiatives.”

Muse Spark is the first model developed by Meta Superintelligence Labs, established last year after CEO Mark Zuckerberg expressed dissatisfaction with the pace of Meta’s advancement, particularly with its Llama models, which were falling behind OpenAI’s ChatGPT and Anthropic’s Claude. To spearhead Meta Superintelligence Labs, Meta brought in Alexandr Wang, co-founder and former CEO of Scale AI, and invested $14.3 billion in the data labeling firm for a 49% stake.

Now, Zuckerberg is eager to determine whether his restructured AI team can attract users.

Muse Spark is currently accessible on the web and through the Meta AI app, with expectations of gradual improvements. The company is set to introduce a “Contemplating” mode that enables it to address more complicated issues. Meta’s model employs multiple AI agents concurrently to tackle the same challenge, which is claimed to yield quicker results for its Contemplating mode.

“To allow for more time spent reasoning without significantly increasing latency, we can enhance the number of parallel agents that collaborate on difficult challenges,” the company stated.

Historically, Meta’s rivals have placed their more advanced models behind a paywall. It remains uncertain whether Meta will adopt a similar approach.

Nevertheless, the company has embraced a trend prevalent in the AI sector. In its blog update, Meta noted that Muse Spark could assist users with health-related inquiries, a field that its competitors are also exploring.

Meta’s venture into health — and even the basic requirement of logging into Muse Spark — might raise privacy issues. Users will need to sign in with an existing Meta account, such as Facebook or Instagram, to access the service. While Meta does not explicitly state that personal data from Facebook or Instagram accounts will be utilized by the AI, it is probable, given that Meta typically relies on public user data for training and has marketed Muse Spark as a personal superintelligence tool.

Additionally, Meta indicated that Muse Spark excels in addressing visual STEM queries, potentially leading to “interactive experiences such as creating entertaining mini-games or troubleshooting household appliances.”

Beyond its investment in Scale AI and the hiring of Wang, Zuckerberg’s company has attracted researchers from OpenAI, Anthropic, and Google. The takeaway: If Meta intends to be a genuine contender in the AI landscape, the time to act is now.

“Looking ahead, we plan to roll out increasingly sophisticated models that advance the limits of intelligence and capabilities, including new open-source models,” Zuckerberg shared on Threads. “We are developing products that not only answer your queries but also function as agents that perform tasks on your behalf.”