Meta Corporation has introduced the AI model Muse Spark, developed by the new team of researchers at Meta Superintelligence Labs (MSL).
Today we’re introducing Muse Spark, our most powerful model yet, giving you a faster and smarter Meta AI.
Muse Spark currently powers the Meta AI app and website and will be rolling out to @whatsapp, @Instagram, @facebook, @messenger, and AI glasses in the coming weeks.…
— Meta Newsroom (@MetaNewsroom) April 8, 2026
This is the first LLM since the company’s CEO, Mark Zuckerberg, embarked on a multi-billion-dollar reorganization of the AI division to reduce the gap with competitors.
The new department is led by Alexander Wang. The head of the corporation invited him as part of investment agreements with Scale AI worth $14 billion. Since then, the corporation has directed substantial funds towards attracting specialized professionals and developing infrastructure.
Since then, the entrepreneur has spent billions of dollars attracting talent in the field of artificial intelligence and allocated tens of billions on infrastructure.
Lagging Behind Competitors
Muse Spark will form the basis of the Meta AI chatbot. The model is closed, unlike the company’s previous products based on open-source code.
The LLM lags behind ChatGPT, Claude, and Gemini in some areas, but the project is in its early stages. The corporation emphasized that the model represents only the “first stage of the journey,” with more powerful solutions under development.
Meta focuses Muse Spark on efficiency and “competitive characteristics” in performing various tasks.
The neural network was created over nine months and is viewed by management as a component of the updated AI strategy. Previously, the firm focused on open-source Llama models.
Muse Spark was trained using several third-party open-source tools like Qwen from the Chinese tech giant Alibaba, as well as solutions from OpenAI and Google.
“Like other industry players, Meta uses distillation with strict safety measures,” a company representative stated.
Modes of Thought
Muse Spark offers several levels of operation: Instant, Thinking, and Contemplating. The latter provides responses at a scientific level. Its integration into the Meta AI app and web version will occur gradually.
Contemplating uses a group of agents for “parallel thinking.” Users can switch between modes depending on the specifics of the request.
Additionally, Meta is testing new monetization methods, offering third-party developers access to the core Muse Spark technology via API.
The company announced that a special “Shopping” mode will be added to the updated version of Meta AI, assisting users in selecting and purchasing clothing.
In February, Meta introduced the AI feature Dear Algo, allowing Threads users to personalize content recommendation algorithms.