Meta has officially stepped into the global race for artificial superintelligence with the launch of Muse Spark, its first major AI model since bringing in Scale AI’s Alexandr Wang nine months ago.12 The release marks a dramatic turning point for the social media giant, which has spent billions rebuilding its entire AI infrastructure from scratch.
Over the last nine months, Meta Superintelligence Labs rebuilt the company’s AI stack from the ground up, moving faster than any development cycle the company has run before.23 This fresh start comes after the disappointing launch of Llama 4 in April 2025, which failed to impress developers and left Meta trailing its rivals.
New Leadership Drives Complete Rebuild
In June 2025, Meta spent $14.3 billion to acquire a 49% nonvoting stake in Scale AI and brought in its cofounder and CEO, Alexandr Wang, as Meta’s first-ever chief AI officer, with Meta announcing it will invest $14.3 billion in Scale AI, which will be overseen by Meta Superintelligence Labs.45
Wang now leads the newly formed Meta Superintelligence Labs, a research division tasked with building what CEO Mark Zuckerberg calls personal superintelligence.
Wang and Zuckerberg went on a talent acquisition spree, offering AI researchers at rival AI labs pay packages that reportedly climbed into the hundreds of millions of dollars when equity was included.4 The aggressive hiring push pulled researchers from OpenAI, Anthropic, and Google DeepMind. Meta said its AI-related capital expenditures in 2026 will be between $115 billion and $135 billion, or nearly twice its capex last year.2
The Scale AI deal shook the AI data industry. Other companies including OpenAI and Google began winding down their relationships with Scale over concerns that Meta could gain access to proprietary training secrets.
Meta Muse Spark artificial intelligence model announcement
What Makes Muse Spark Different
This initial model is small and fast by design, yet capable enough to reason through complex questions in science, math, and health.3
The model is exposed as two different modes on meta.ai: Instant and Thinking, with Meta promising a Contemplating mode in the future which they say will offer much longer reasoning time and should behave more like Gemini Deep Think or GPT-5.4 Pro.6
The Contemplating mode uses something Meta calls multi-agent orchestration. For that mode, Muse Spark uses a squad of AI agents to help reason in parallel, helping it compete with the extreme reasoning modes of frontier models such as Gemini Deep Think and GPT Pro.2
Key Capabilities:
- Multimodal perception for analyzing images and charts
- Visual coding to create mini-games and custom websites
- Health guidance powered by collaboration with over 1,000 physicians
- Shopping mode that draws styling inspiration from Instagram and Facebook creators
Meta said improved AI training techniques along with rebuilt technology infrastructure has enabled the company to create smaller AI models that are as capable as its older midsize Llama 4 variant for an order of magnitude less compute.26 This efficiency gain represents a major technical achievement for the team.
Performance Against Rivals
The model is, according to benchmark tests that Meta published, competitive with leading AI models from OpenAI, Anthropic, and Google across many tasks, although it does not surpass them across the board.4
Meta’s internal benchmarks show Muse Spark ranking behind only Gemini 3.1 Pro and GPT-5.4 in several categories. Artificial Analysis scored Meta Spark at 52, behind only Gemini 3.1 Pro, GPT-5.4, and Claude Opus 4.6.6 Last year’s Llama 4 models scored just 18 and 13 on the same scale.
However, questions remain about the accuracy of these numbers. In the past, Meta has been caught manipulating the published benchmark results of an AI model to make it appear more capable than the version available to most users.4
Meta continues to invest in areas with current performance gaps, specifically long-horizon agentic systems and coding workflows.24 The company acknowledges it still trails rivals in coding tasks and complex multi-step planning.
Strategic Shift From Open Source
The launch of Muse Spark represents a major departure from Meta’s previous philosophy. The company had been taking an open-source approach to AI with its Llama family of models.2
Muse Spark will power the Meta AI chatbot, and in a pivot from the company’s prior open-source strategy, is a closed model, meaning its design and code won’t be made public, with the company saying there is hope to open-source future versions of the model.27
This decision marks a strategic calculation. Building frontier AI models now costs hundreds of billions of dollars in infrastructure. By keeping Muse Spark proprietary, Meta gains more control over how the technology is used and opens new revenue opportunities.
Meta is also experimenting with a new AI model revenue stream by offering third-party developers access to Muse Spark’s underlying technology via an API, with currently only unspecified select partners able to access the AI model’s private API preview, but it plans to eventually offer paid API access to a wider audience at a later date.2
Where You Can Use It
Muse Spark currently powers the Meta AI app and website, and will be rolling out to WhatsApp, Instagram, Facebook, Messenger, and AI glasses in the coming weeks.3
The model brings practical features users can try immediately. Snap a photo of an airport snack shelf and Meta AI can identify and rank the snacks with the most protein.3 The system can also help plan family trips by launching multiple AI agents in parallel to draft itineraries, compare destinations, and find activities simultaneously.
With Muse Spark, Meta AI is now able to help you navigate health questions with more detailed responses, including some questions involving images and charts, with the company working with a team of physicians to develop the model’s ability to provide helpful information on common health questions and concerns.38
Privacy concerns have already emerged. Muse Spark users will need to log in with an existing Meta account such as Facebook or Instagram in order to use it, and while Meta doesn’t explicitly say that personal information from a Facebook or Instagram account will be used by the AI, it is likely, considering that Meta generally trains on public user data and the company has positioned Muse Spark as a personal superintelligence product.9
Meta’s stock popped 6.5% on Wednesday, rallying alongside the rest of the market, with Meta Platforms shares rising nearly 9% on April 8, 2026, following the announcement.210 Wall Street appears optimistic that the massive infrastructure investments are finally paying off with competitive products.
The launch of Muse Spark signals that Meta is no longer content to sit on the sidelines of the AI revolution. After months of silence while rivals shipped reasoning models and consumer products, the company has returned with a legitimate contender. Whether this closed approach will win over the developers who once championed Llama’s openness remains the biggest question. But for now, Meta has proven it can compete at the frontier.
What do you think about Meta’s decision to abandon open source for its most powerful AI models? Does the efficiency and capability of Muse Spark justify the closed approach, or should the company have stayed true to its original philosophy? Drop your thoughts in the comments below.
