Meta Unveils Muse Spark AI Model, Superintelligence Team's First Release
- tech360.tv
- 2 hours ago
- 3 min read
Meta Platforms has unveiled Muse Spark, its initial artificial intelligence (AI) model developed by a newly assembled "superintelligence" team. Shares of the company extended gains, trading up nearly 7% following the announcement. This new model represents Meta's efforts to compete with rivals in the AI sector.
The company assembled its superintelligence team last year, hiring Chief Executive Officer Alex Wang from Scale AI under a $14.3 billion deal. Some engineers were offered pay packages of hundreds of millions of dollars to staff the team, aiming to advance Meta’s position in AI development.
"Superintelligence" refers to AI machines that could potentially surpass human thought capabilities. Muse Spark is the first in a new series of models internally known as Avocado, developed by this specialised team.
Muse Spark will initially be available on the Meta AI application and website. The company stated that the model will replace existing Llama models that power chatbots on WhatsApp, Instagram, Facebook, and Meta’s smart glasses in the coming weeks.
Meta did not disclose Muse Spark’s size, a metric typically used to compare an AI system’s computing power with competitors. Unlike previous open releases of its Llama models, Meta shared only a "private preview" of Muse Spark with unnamed partners.
According to a company blog post, "This initial model is small and fast by design, yet capable enough to reason through complex questions in science, math, and health. It is a powerful foundation, and the next generation is already in development."
Independent evaluations of Muse Spark’s performance showed it catching up with top models from market leaders Google, OpenAI, and Anthropic in some areas, such as language and visual understanding. However, the model lagged in coding and abstract reasoning, tying for fourth place on a broad index of AI tests compiled by Artificial Analysis.
Meta Chief Executive Officer Mark Zuckerberg tempered expectations in January. He stated the team’s first models "will be good but, more importantly, will show the rapid trajectory that we’re on." Zuckerberg added, "I expect us to steadily push the frontier over the course of the year as we continue to release new models."
Wang, who leads the superintelligence team, acknowledged that "there are certainly rough edges we will polish over time in model behavior." He confirmed that larger versions of Muse Spark are under development, with plans to release some openly.
Meta also provided insights into its monetisation strategy, teasing shopping features embedded within its Meta AI chatbot. These features aim to direct users to purchasable products.
The company is betting that applying AI to everyday personal tasks will increase engagement among its more than 3.5 billion users across its social media platforms. This strategy could provide Meta an advantage over rivals with smaller user bases.
Muse Spark can assist users with tasks such as estimating calories in a meal from a photo or superimposing an image of a mug on a shelf. An additional Contemplating Mode runs multiple agents simultaneously to boost reasoning power.
This mode would allow Muse Spark to take on the extended thinking modes of Google’s Gemini Deep Think and OpenAI’s GPT Pro. Users could leverage this for tasks like efficiently planning a family vacation, with one agent drafting an itinerary while another searches for kid-friendly activities.
Meta Platforms launched Muse Spark, the first AI model from its superintelligence team.
The model will initially be available on the Meta AI app and website, later replacing existing Llama models across Meta's platforms.
Independent evaluations show Muse Spark performs well in language and visual understanding but lags in coding and abstract reasoning.
Source: REUTERS
