Meta Unveils Muse Spark to Strengthen AI Ambitions
Meta Platforms has introduced Muse Spark, marking the first artificial intelligence model developed by a specialised team formed last year to compete more effectively in the intensifying AI race. Following the announcement, the company’s shares climbed significantly, rising by nearly seven percent.
The launch reflects growing pressure on major technology firms to justify their substantial investments in artificial intelligence. For Meta, the stakes are particularly high after committing heavily to talent and infrastructure in an effort to regain momentum in the field. The company previously faced disappointment with its earlier Llama 4 models, which did not meet expectations.
New Direction for Meta’s AI Strategy
Muse Spark represents the beginning of a new internal model series known as Avocado. Unlike earlier releases, the model will initially be available only through Meta’s AI app and website. However, the company plans to integrate it across its platforms in the coming weeks, replacing existing Llama-based systems used in chatbots on WhatsApp, Instagram, Facebook, and its smart glasses.
Notably, Meta has shifted away from its previous approach of openly releasing models. Instead, it is offering Muse Spark through a limited private preview to selected partners. Furthermore, the company has not disclosed the model’s size, a common benchmark used to evaluate AI capability.
According to Meta, the model has been designed to be both efficient and capable. It can handle complex queries in areas such as science, mathematics, and health, while maintaining speed and responsiveness. The company also indicated that more advanced versions are already under development.
Performance and Early Limitations
Independent evaluations suggest that Muse Spark performs competitively in areas like language processing and visual understanding. However, it still trails leading systems in coding and abstract reasoning tasks. Overall, it secured a joint fourth-place ranking on a comprehensive AI performance index.
Meta leadership has acknowledged that the model is still evolving. Early expectations were deliberately measured, with emphasis placed on long-term progress rather than immediate dominance. The development team has also recognised existing limitations, noting that improvements will be made over time.
Monetisation and Practical Applications
The release of Muse Spark also signals a clearer commercial direction. Meta has begun testing shopping features within its AI chatbot, allowing users to discover and purchase products directly through the platform. This approach aligns with the company’s broader strategy of embedding AI into everyday digital experiences.
With access to billions of users across its platforms, Meta aims to increase engagement by offering practical AI-driven tools. These include features such as estimating calorie content from food images and visualising objects within real-world settings.
Additionally, a specialised Contemplating Mode enables the system to run multiple agents simultaneously, enhancing its reasoning capabilities. This allows users to complete more complex tasks, such as planning trips, by dividing responsibilities between different AI processes.
With inputs from Reuters