Metaが初の独自AIモデル「Muse Spark」をリリース

3 Min ReadSocial media giant Meta on Wednesday introduced its first AI model from its Superintelligence Lab, almost a year after the team was created, with CEO Mark Zuckerberg revealing that the company will focus on personal superintelligence.Muse Spark is the first in the Muse family of models, with multimodal reasoning capabilities, support for tool use, visual chain of thought and multi-agent orchestration. The model is available at meta.ai and the Meta AI app. Meta is also opening a private API preview to some users.The model boasts capabilities in multimodal perception, reasoning, health and agentic tasks. However, Meta acknowledged that it is investing in performance gaps in areas such as coding and agentic systems that solve multi-step tasks. Despite these shortcomings, Meta also released Contemplating mode, which enables Muse Spark to compete with the reasoning modes of models like Gemini Deep Think and GPT Pro from OpenAI.Related:The Real AI Shift Isn’t New Models. It’s Control.With Muse Spark, Meta has created a model that can understand users' personal worlds, such as their personal health. The model targets personal AI use, with the social media giant focusing on consumers and everyday users in its health applications. However, the social media giant is also looking to be a technology leader in AI and providing a personal use model could help it get there.Meta's Strategy"Meta always has a clear strategy when it comes to their product launches," said Lian Jye Su, an analyst at Omdia, a division of Informa TechTarget. "It's designed for our day-to-day stuff and the interaction on Meta's app ecosystem," Su said. "It probably will work pretty well within that ecosystem at this point."Moreover, enterprises that already interact with Meta's social media app ecosystem (WhatsApp, Facebook and Instagram) might also start testing and playing with Muse Spark."What Meta probably wants is to get people to get familiar with the AI model, for them to understand its preferences and its capability and strength and limitation," Su said. Too Many ModelsHowever, as the first closed model Meta has released (its open model being the Llama family), there is a danger that the model could be another "me too" model, as it competes with models from Anthropic, OpenAI, Claude and others, Su added. But Meta's advantage is its large data repository. Although Meta does not have as many productivity tools as Microsoft or Google, it still has enough data to influence how AI is deployed or used daily, Su continued."It's going to have a risk to become a 'me too' product, but at least it is going to be the only AI product that's compatible with the popular apps that Meta likes -- WhatsApp, Facebook or Instagram," he said.Related:OpenAI GPT-5.4-Cyber is More Open Than Claude MythosHe added that the new Muse models will be a fresh start for the vendor, which previously released open source models, and that targeting a personal audience is not necessarily a weakness and could extend Meta's reach to enterprises as well."If more people are using their platform because of AI, it benefits all the other enterprises that are advertising on their platform or using their platform to reach out to more users," Su said.About the AuthorNews Writer, AI BusinessEsther Shittu brings four years of expertise covering artificial intelligence technologies and industry trends. As co-host of the "Targeting AI" podcast, she talks to thought leaders and practitioners exploring critical AI developments. Previous to AI Business, she wrote for several publications including the New York Daily News, Bklyner and the Brooklyn Daily Eagle. When she's not diving deep into the world of AI, she spends her time on passion projects and raising her three daughters.