Goodbye, Llama? Meta launches new proprietary AI model Muse Spark — first since Superintelligence Labs' formation

2026-04-08 · Source: VentureBeat · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation · Depth: Advanced, medium

Summary

Meta has launched Muse Spark, a new proprietary AI model developed by its Meta Superintelligence Labs (MSL) division, led by Alexandr Wang. This release follows a significant overhaul of Meta's AI operations in mid-2025 after the mixed reception of Llama 4. Muse Spark is described as Meta's most powerful model, featuring tool-use, visual chain of thought, and multi-agent orchestration, and is intended to be the foundation for "personal superintelligence." It is a natively multimodal reasoning model, integrating visual information from the ground up, and introduces a "Contemplating" mode for parallel reasoning. Benchmarks show Muse Spark achieving an Artificial Analysis Intelligence Index score of 52, placing it among the top global models, and demonstrating strong performance in multimodal reasoning, health, and token efficiency, using significantly less compute than Llama 4 Maverick. The model is currently proprietary, available via the Meta AI app, website, and a private API preview.

Key takeaway

For CTOs and VPs of Engineering evaluating foundational models, Muse Spark signals Meta's pivot to proprietary, high-performance multimodal AI. While its proprietary nature departs from the Llama series' open-source accessibility, its strong benchmark performance in visual reasoning and health, coupled with significant token efficiency, makes it a compelling option for applications requiring advanced perception and reduced operational costs. Consider its private API preview for specific use cases, especially if your strategy aligns with personal superintelligence or agentic systems, but be aware of its current limitations in abstract reasoning and agentic workflows.

Key insights

Muse Spark marks Meta's return to frontier AI, emphasizing multimodal reasoning and efficiency over its prior open-source Llama strategy.

Principles

• Multimodal integration enhances reasoning.
• Efficiency through "thought compression" is key.
• Agentic systems require refined real-world execution.

Method

Muse Spark employs a natively multimodal architecture, integrating visual information directly, and utilizes a "Contemplating" mode for orchestrating multiple sub-agents to reason in parallel, alongside "thought compression" to reduce compute.

In practice

• Deploy Muse Spark for advanced visual reasoning tasks.
• Utilize its health sector capabilities for medical applications.
• Explore its "Shopping Mode" for personalized e-commerce.

Topics

• Muse Spark
• Meta Superintelligence Labs
• Multimodal Reasoning
• Visual Chain of Thought
• AI Benchmarking

Best for: Investor, CTO, VP of Engineering/Data, AI Scientist, Director of AI/ML, Tech Journalist

Related on AIssential

• Introducing Muse Spark, the first in the Muse family of models developed by Meta Superintelligence Labs. Muse Spark is a natively multimodal reasoning model with support for tool-use, visual chain of thought, and multi-agent orchestration. Muse Spark is avai - x.com — X.com requires JavaScript and may conflict with privacy extensions for full functionality.
• Meta Llama 3.1 405B May Spark a Tectonic Shift in the LLM Market Adoption — Meta's Llama 3.1 405B is the first open model to achieve frontier-level performance, rivaling top proprietary LLMs.
• Meta debuts new AI model in first test of costly ‘superintelligence’ team — Meta's Muse Spark model aims to compete with leading AI systems, focusing on practical applications and user engagement.
• All of AI's New Models and Tools — New AI models and tools emphasize multimodal reasoning, open-source access, and scalable agentic capabilities.
• Meta is back in the LLM game after a year-long break — Meta's Muse Spark release signals a renewed, resource-intensive push into top-tier LLMs, despite past failures and ongoing post-training challenges.
• Meta AI Muse Spark IS INCREDIBLE! Powerful Coding & Multimodal Model! (Fully Tested) — Muse Spark is Meta's new multimodal AI, excelling in visual reasoning, coding, and agentic workflows.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by VentureBeat.