In early April, Meta revealed Muse Sparkits latest AI model and first major release under AI chief Alexander Wang. The model performs competitively in some benchmarks, but it still holds up OpenAI‘s GPT-5.4 Pro and GoogleGemini 3.1 Pro, prompting questions about whether the Meta is lagging behind. Wang believes the framing makes no sense.
“The new Muse Spark model we released is not at the level of the leading edge models,” Wang said during an on-stage interview at the Bloomberg Tech Summit in San Francisco yesterday (June 4). “But we believe it’s a very exciting data point in the trajectory, and we expect the next models we release to be very competitive with the leading models in the world.”
He described the Muse Spark as an “appetizer.” Asked when the entry would arrive, Wang replied: “We are cooking it up. We are seeing very exciting and promising results in its training process now.”
Muse Spark marks a change for Meta. It’s the company’s first proprietary model and is only built into Meta’s products, rather than being openly released like previous systems were. The model is designed to handle text, images, video and audio and to support more complex, multi-step tasks, including content-related shopping features on Instagram and Facebook.
The release follows a difficult stretch for Meta’s AI efforts. Llama 4, launched in April 2025, was widely criticized. Two months later, Mark Zuckerberg hired Wang to lead his newly formed Superintelligence Labs and reset its strategy.
Wang said the group is focused on scaling: expanding data, computing power and research to drive improvements. And Muse Spark sits early in that process.
The obstacle to the border is not money, Wang said. “It’s about continuing to scale data, computing … and continuing to scale with research. All the labs are dramatically scaling up their models, and we’re on a much faster trajectory because we’ve done all this work over the last year.”
Meta is supporting this approach with heavy overhead. The company expects capital spending of $125 billion to $145 billion in 2026, up from $72.2 billion in 2025, and is targeting more than 1.3 million GPUs and roughly one gigawatt of AI computing capacity.
The move to a closed model also reflects security concerns. During development, the Muse Spark triggered internal alarms, including about potential biohazards.
“When the company releases a model in a product, we have many ways to mitigate some of these risks,” he said. “It’s much harder to do that when you open source the model.”
Meta hasn’t completely abandoned open source AI and continues to develop models it deems safe to release. Whether its Llama brand will continue remains undecided. “We have exciting debates about internal branding,” Wang said, “and nothing to share now.”
Wang said Muse Spark’s strengths are in multimodal capabilities, health-related applications and creative coding — such as generating simple games or digital tools. These areas support Meta’s broader push into AI agents.
The company is “doubling down” on agents, aiming to build what Wang called “the best personal agents for everyone around the world.” He said he uses such tools himself to manage his health and stay in touch with friends.
The push comes along with major internal changes. In May, Meta announced about 8,000 layoffs and reassigned about 7,000 others to AI-focused roles as part of a broader reorganization.
“It’s incredibly difficult to say goodbye to your teammates,” Wang said. “We don’t take either lightly.”
