Meta launches Muse Spark, its first AI model from Superintelligence Labs, in a major reboot of its AI strategy
Built from scratch by the team Zuckerberg assembled after being disappointed by Llama 4, Muse Spark benchmarks favorably against models from OpenAI, Anthropic, and Google in early evaluations.
Meta released Muse Spark on April 8, marking the debut of a substantially rebuilt AI strategy following CEO Mark Zuckerberg's decision last year to overhaul the company's AI efforts from the ground up. The model is the first release from Meta Superintelligence Labs — the unit Zuckerberg created after growing frustrated with Llama 4's performance relative to OpenAI's ChatGPT and Anthropic's Claude — and is now available on the web and the Meta AI mobile app.
Muse Spark is inherently multimodal, designed to process and reason across images, audio, video, and text. Meta says it performs particularly well on visual STEM problems, health-related queries, and coding tasks, and can spawn multiple parallel subagents to attack complex problems simultaneously.
Zuckerberg wrote 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."
The model was developed under the leadership of Alexandr Wang, the former Scale AI CEO who joined Meta after the company invested $14.3 billion for a 49 percent stake in Scale. Meta has also recruited heavily from OpenAI, Anthropic, and Google, offering packages reportedly worth hundreds of millions of dollars to top researchers.
The company describes Muse Spark as "the first step on our scaling ladder," indicating it plans to release progressively more capable models.
Early independent benchmarking from Artificial Analysis, which had early access, placed Muse Spark 52nd on its Artificial Index — within the top tier of evaluated models. Meta positioned the launch as the starting point of a serious effort to compete with the leading U.S. AI labs, with future open-source releases promised as the company scales up its capabilities.
Read the original reporting at TechCrunch.
