Meta’s New AI Model Gives Mark Zuckerberg a Seat at the Big Kid’s Table
Meta's closed-source Muse Spark scores 52 on the AI Index, placing it in the top 5 models benchmarked.
Meta has unveiled Muse Spark, a new flagship AI model developed by its Meta Intelligence Labs division, marking a significant strategic pivot. Unlike its open-source Llama predecessors, Muse Spark is being released as a closed-source model, available only through meta.ai and the Meta AI app. The company reports that the model outperforms the latest offerings from OpenAI, Anthropic, Google, and xAI on certain tasks. Independent benchmarking firm Artificial Analysis, which had early access, scored Muse Spark at 52 on its Intelligence Index, ranking it within the top five models it has ever tested.
Technically, Muse Spark is a natively multimodal model, trained from scratch to handle text, images, audio, and video with advanced reasoning and strong coding capabilities. A key focus area is medical advice; Meta collaborated with over 1,000 physicians to curate specialized training data for this domain. The launch follows a massive investment overhaul by CEO Mark Zuckerberg, who has spent billions acquiring AI startups and recruiting top talent—including Scale AI's CEO Alexandr Wang—to reboot the company's AI ambitions. Muse Spark is framed as the foundational step in Meta's 'Advanced AI Scaling Framework,' a roadmap aimed at safely developing AI that far surpasses human intelligence.
- Muse Spark scores 52 on the Artificial Analysis Intelligence Index, placing it in the top 5 benchmarked models globally.
- The model is closed-source, a departure from Meta's previous open-source strategy with its Llama series of models.
- It was developed with input from over 1,000 physicians specifically to enhance its medical reasoning and advice capabilities.
Why It Matters
This closed-source, high-performing model signals Meta's aggressive push to compete directly with OpenAI and Google in the commercial AI race.