Meta has unveiled the first tangible result of Mark Zuckerberg's sweeping AI overhaul, aimed at closing the gap with industry leaders. Developed by Meta Superintelligence Labs (MSL), Muse Spark is the product of nine months of intensive work by the team led by Alexandr Wang, who joined Meta following the company's $14.3 billion investment in Scale AI. "Nine months ago we rebuilt our AI stack from scratch. New infrastructure, new architecture, new data pipelines," Wang announced on social media, framing the release as a foundational reset rather than an incremental upgrade.
Natively multimodal by design, Muse Spark accepts voice, text, and image inputs, though it currently produces text-only output. Its standout feature is the new "Contemplating" mode, which orchestrates multiple sub-agents to reason in parallel — a direct response to Google's Gemini Deep Think and OpenAI's GPT-5.4 Pro extended reasoning capabilities. Meta claims the model achieves this reasoning capacity using more than ten times less compute than Llama 4 Maverick, thanks to a training technique called "thought compression," which penalizes excessive thinking time during reinforcement learning to deliver efficiency without sacrificing accuracy.
The benchmark results paint a mixed picture. Muse Spark ranks fourth on the Artificial Analysis Intelligence Index v4.0 with a score of 52, behind Gemini 3.1 Pro Preview and GPT-5.4 (both at 57) and Claude Opus 4.6 (53). On the abstract reasoning benchmark ARC AGI 2, Muse Spark trails significantly with a score of 42.5, but it leads the pack on CharXiv Reasoning — a chart and figure understanding evaluation — with a score of 86.4. On HealthBench Hard, the model scored 42.8%, outperforming its rivals thanks to training data curated in collaboration with more than 1,000 physicians.
The detail drawing the most attention from the developer community, however, is that Muse Spark is closed source. As one of the pioneers of open-source AI development through its Llama series, Meta is breaking with tradition. While the company has stated it hopes to release future versions under an open-source license, the move signals a more strategic stance toward protecting competitive advantage. Muse Spark is now powering queries in the Meta AI app and the Meta.ai website, with rollouts planned across WhatsApp, Instagram, Facebook, and Messenger in the coming weeks. Zuckerberg emphasized that the model is "particularly strong in areas related to personal superintelligence like visual understanding, health, social content, shopping, games, and more."