Back
Meta AI BlogLaunchMeta AI Blog2026-04-10

Meta Launches Muse Spark — The First Model From Its New Superintelligence Labs

Meta's new Superintelligence Labs has shipped Muse Spark, the company's most capable model yet — natively multimodal with parallel agent reasoning. Rolling out across Facebook, Instagram, WhatsApp, and Ray-Ban glasses.

Original source

Meta has quietly become the most important AI company you're not paying attention to. Muse Spark, released on April 8, is the first model to come out of Meta Superintelligence Labs (MSL) — the unit founded by Scale AI's Alexandr Wang earlier this year. And it's a significant step change from everything Meta has shipped before.

Muse Spark is natively multimodal: it accepts text, images, and voice as inputs and generates text and code responses. More importantly, it uses a parallel agent reasoning architecture — before responding, it dispatches multiple internal reasoning threads simultaneously and synthesizes the best result. This is Meta's answer to OpenAI's o3 and Gemini's Deep Think reasoning modes, and according to early internal benchmarks, it's competitive on complex multi-step reasoning tasks.

The model is rolling out inside Meta's consumer products — Facebook, Instagram, WhatsApp, Messenger, and Meta AI on Ray-Ban smart glasses. That distribution scale is something no other frontier lab can match: Muse Spark will reach billions of users without a single API key. For developers, a limited API preview is expected in Q3 via Meta AI Studio.

What's notable is the speed. Wang only joined Meta in late 2025, and MSL has shipped a frontier-competitive model in under six months. Whether that reflects Meta's pre-existing research foundation or genuine new capability from the new lab remains to be seen — but the result speaks for itself. The era of Meta as a "fast follower" in frontier AI may be over.

Panel Takes

The Builder

The Builder

Developer Perspective

The parallel agent reasoning architecture is genuinely interesting from a systems perspective — it's a different tradeoff than chain-of-thought scaling. The real test will be the API: if Meta opens Muse Spark to developers with competitive pricing, it changes the landscape significantly.

The Skeptic

The Skeptic

Reality Check

Meta has announced 'best model ever' moments before and then quietly let things stagnate. MSL is a fresh coat of paint on a company that still monetizes through surveillance advertising. I want to see independent benchmark results before adjusting my priors on whether this is actually frontier.

The Futurist

The Futurist

Big Picture

Three billion people getting access to a frontier-competitive multimodal AI model embedded in apps they already use every day — that's not a product launch, it's a civilizational event. The AI literacy implications alone are staggering. Meta may have just become the most impactful AI deployment in history.