AI: Meta casts Muse Spark to forget about Llama 4’s flop

Unlike its predecessors, Muse Spark was born with an architecture designed specifically for the company’s products, emulating the tight vertical integration already seen with Google Gemini. The technology supports multimodal input, allowing the user to alternate between an immediate response mode and a more reflective one, called “Thinking”, for solving complex logical tasks. The introduction of sub-agents operating in parallel should also ensure faster and more efficient management of requests, especially in the recognition of texts and images acquired in real time via the camera of wearable devices.

The main battleground of the new series, however, seems to be that of health, where Meta openly challenges the specialized solutions of OpenAI and Anthropic presented at the beginning of the year. Muse Spark promises to navigate complex medical questions by analyzing graphs and images, such as the caloric estimate of a meal, although the accuracy of such tools remains a subject of debate due to risks related to misinformation. Looking ahead, the group intends to use the content shared on Threads and Instagram to personalize the recommendations provided by the AI, while maintaining the intention to release open-source versions of future more advanced models of the Muse family.