Meta has released ‘Muse Spark’, its new LLM designed to power a smarter and faster Meta AI and soon cite recommendations and content people share across Instagram, Facebook, and Threads.
It’s the first AI model to be released under the auspices of the tech giants’ chief AI officer and apparent tech wunderkind Alexandr Wang, who was hired mid-way through 2025.
The model, built by the recently formed Meta Superintelligence Labs is the first step in Meta’s plan to overhaul its AI efforts.
As Meta puts it, Muse Spark is a “natively multimodal reasoning model with support for tool-use, visual chain of thought, and multi-agent orchestration,” and is built to be integrated directly into other Meta’s products.
Meta said Muse Spark is the first step in its quest to produce its own “personal superintelligence model” that understands a personalised human world world.
The AI model, according to Meta, offers competitive performance in multimodal perception, reasoning, health, and agentic tasks. Yes, you read that right, one application of the personal superintelligence model is to help people learn about and improve their health.
To improve Muse Spark’s health reasoning capabilities, Meta explained it collaborated with over 1,000 physicians to curate training data that enables more factual and comprehensive responses. Muse Spark can generate interactive displays that unpack and explain health information such as the nutritional content of various foods or muscles activated during exercise.
Superintelligence Labs
Meta’s Superintelligence Labs, was created last year out of CEO Mark Zuckerberg’s frustration at the progress (or lack thereof) of Meta’s Llama models as it lagged behind OpenAI’s ChatGPT and Anthropic’s Claude.
As such, Muse Spark boasts a Contemplating mode. According to the tech giant, it “provides significant capability improvements in challenging tasks, achieving 58 per cent in Humanity’s Last Exam and 38 per cent in Frontier Science Research.
The mode also orchestrates multiple agents that reason in parallel. This is so that Meta and Muse Sparks can compete with the extreme reasoning modes of frontier models such as Gemini Deep Think and GPT Pro.
Superintelligence labs is ran by Scale AI ex-CEO Wang and former GitHub CEO Nat Friedman.
Wang said it will be the start of a new Muse family of models, which has raised questions about what will happen to the ongoing development of the Llama family—Meta’s family of open-source generative AI models.
To support further scaling, the tech giant and investing heavily into research, model training, infrastructure and its Hyperion data centre—created to generate the computing power necessary to achieve Meta’s artificial intelligence ambitions.
Zuckerberg disclosed the tech giants next steps on Threads, “looking ahead, we plan to release increasingly advanced models that push the frontier of intelligence and capabilities, including new open source models. We are building products that don’t just answer your questions but act as agents that do things for you.”
