Meta has announced Muse Spark, the first in a new series of large language models developed by its Meta Superintelligence Labs, as it pushes towards building more advanced and personalised artificial intelligence tools.
The company said Muse Spark is designed to power a faster and more capable Meta AI assistant, with plans to introduce new features that incorporate recommendations and content shared across Instagram, Facebook and Threads.
The model will underpin the next phase of Meta’s AI ambitions, with larger and more advanced systems already in development. Muse Spark currently powers the Meta AI app and website and is expected to roll out to WhatsApp, Instagram, Facebook, Messenger and AI-enabled glasses in the coming weeks. It will also be made available in private preview to selected partners via an API.
Meta described the launch as a step towards what it calls “personal superintelligence” — an assistant designed to help users with everyday tasks and more complex challenges.
A new generation of AI models
According to the company, Meta Superintelligence Labs has rebuilt its AI systems over the past nine months, adopting a more structured approach to scaling its models. Muse Spark is the first in the new “Muse” series, designed to incrementally improve performance with each iteration.
Despite being smaller and faster by design, the model is capable of handling complex queries across areas such as science, mathematics, and health, forming the foundation for more advanced versions already in development.
Upgraded Meta AI experience
The introduction of Muse Spark also brings upgrades to the Meta AI app and web platform, including a redesigned interface and enhanced functionality. Users can now switch between different modes depending on the complexity of their query, while the system can deploy multiple AI agents simultaneously to handle different aspects of a task.
For example, when planning a trip, separate agents can generate an itinerary, compare destinations, and suggest activities at the same time, delivering faster and more detailed results.
Enhanced visual and multimodal capabilities
A key feature of Muse Spark is its multimodal capability, allowing Meta AI to process and interpret images as well as text. This enables users to take photos and receive contextual insights, such as identifying products or comparing items in real time.
Meta said this capability is particularly useful in areas such as health, where users increasingly turn to AI for guidance. The model has been developed with input from medical professionals to improve responses to common health-related questions, including those involving images and charts.
New tools for creativity and discovery
Muse Spark also introduces enhanced creative tools, including visual coding features that allow users to generate websites, dashboards and simple games from text prompts.
In addition, Meta AI now includes a shopping mode that draws on content from creators and communities across its platforms, helping users discover fashion, home styling ideas and gift suggestions.
The assistant can also provide contextual insights on locations and trending topics by surfacing relevant public posts and discussions, offering users a broader perspective based on community activity.
Rollout and future plans
The upgraded Meta AI experience is being rolled out in the United States, with plans to expand to additional countries and platforms in the coming weeks. Meta said it also intends to open access to the underlying technology, beginning with a private preview for selected partners.
Looking ahead, the company said future updates will deliver richer and more visual responses, integrating content such as videos, photos and posts directly into answers while crediting original creators.
Meta added that it will continue to strengthen safeguards around safety and privacy as its AI systems evolve, as part of its broader ambition to build more intuitive and context-aware digital assistants.
No comments:
Post a Comment