Meta releases Muse Spark, first model from Meta Superintelligence Lab to “power a smarter and faster” Meta AI

What Meta unveiled

Meta on Wednesday unveiled Muse Spark, a new AI model it says was “purpose-built” to power a smarter, faster Meta AI across Facebook, Instagram and Threads. The company says the model will deliver more personalised, visual responses — weaving Reels, photos and posts into answers and crediting creators — and will add features such as a shopping mode and deeper health-related guidance after work with more than 1,000 doctors. It has been reported that Meta positions the rollout as an early, product-focused step rather than a show-stopping leap past rivals.

Not an open Llama, and not quite the frontier

Unlike Meta’s previous Llama family, Muse Spark is a smaller, closed model offered initially via a private preview to select partners; Meta says it hopes to open-source future versions. The company shared evaluations suggesting Muse Spark outperformed some competitors on select reasoning and multimodal benchmarks — a claim that, it has been reported, comes from Meta’s own testing. Meta also conceded gaps remain, notably in long-horizon agentic systems and coding workflows.

Leadership, spending and the stakes

This release is the first from the Meta Superintelligence Lab, a reorganised unit led by Alexandr Wang after a costly push by Mark Zuckerberg to bulk up AI talent and infrastructure. It has been reported that the move followed heavy investment in data centres and the hiring spree that included a large investment tied to Wang’s former company. Investors have been impatient; after Llama 4 disappointed expectations last year, Zuckerberg shuffled teams and doubled down. The emotional moment? Meta is no longer experimenting on the side — it’s onstage, with billions on the line.

Early step, not finish line

Muse Spark is an early indicator of how Meta intends to fold social content into conversational AI and to chase the promise of “personal superintelligence.” Will it be enough to close the gap with OpenAI and Google? Not yet. But the model signals Meta’s shift from open research toys to product-first, competitive play — and the race just got more interesting.

Sources: ft.com