Meta is debuting its first major artificial intelligence model since the costly hiring of Scale AI’s Alexandr Wang nine months ago, as the Facebook parent aims to carve out a niche in a market that’s being dominated by OpenAI, Anthropic and Google.
Dubbed Muse Spark and originally codenamed Avocado, the AI model announced Wednesday is the first from the company’s new Muse series developed by Meta Superintelligence Labs, the AI unit that Wang oversees. Wang joined Meta in June as part of the company’s $14.3 billion investment in Scale AI, where he was CEO.
Meta is desperate to regain momentum in the fiercely competitive AI market following the disappointing debut of its latest open-source models last April. The release failed to captivate developers, leading CEO Mark Zuckerberg to pivot his strategy.
“Over the last nine months, Meta Superintelligence Labs rebuilt our AI stack from the ground up, moving faster than any development cycle we have run before,” Meta said in a blog post on Wednesday. “This initial model is small and fast by design, yet capable enough to reason through complex questions in science, math, and health. It is a powerful foundation, and the next generation is already in development.”
Meta’s stock popped almost 9% on Wednesday, and headed for its sharpest rally since January. The shares gained alongside the rest of the market, which jumped after President Donald Trump said he was suspending Iran attacks for two weeks, sending oil prices tumbling.
Meta isn’t positioning Muse Spark as a top-of-the-line model, but is instead highlighting its efficiency and “competitive performance” on various tasks.
While Meta has used advancements in generative AI and its own investments in the technology to bolster its advertising business and improve efficiencies across the company, it’s yet to crack the AI model market in a significant way, and its top competitors in the space zoom have zoomed ahead. OpenAI and Anthropic are now collectively valued at over $1 trillion, and Google’s Gemini technology and services have gained traction, particularly in the consumer market.
The stakes are massive, as the global generative AI market is estimated to grow more than 40% a year, climbing from about $22 billion in 2025 to almost $325 billion by 2033, according to Grand View Research.
Meanwhile, Meta is ramping up its spending on AI infrastructure, trying to keep up with the other hyperscalers. In its latest earnings report, Meta said its AI-related capital expenditures in 2026 will be between $115 billion and $135 billion, or nearly twice its capex last year.
The new Muse Spark will be proprietary, with the company saying there is “hope to open-source future versions of the model.” The company had been taking an open-source approach to AI with its Llama family of models.
Meta said in a technical blog about the new model that that 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.”
“Muse Spark offers competitive performance in multimodal perception, reasoning, health, and agentic tasks,” Meta said in the post. “We continue to invest in areas with current performance gaps, specifically long-horizon agentic systems and coding workflows.”
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. Currently, only unspecified “select partners” can access the AI model’s “private API preview,” but Meta said it plans to eventually offer paid API access to a wider audience at a later date.
The new model now powers the company’s digital assistant in the standalone Meta AI app and desktop website. Muse Spark will debut in the coming weeks inside Facebook, Instagram, WhatsApp and Messenger, as well as in the company’s Ray-Ban Meta AI glasses. Meta also plans for Muse Spark to eventually power the company’s Vibes AI video feature in the Meta AI app. That service currently uses AI models from third-parties like Black Forest Labs.
Alexandr Wang speaks on CNBC’s “Squawk Box” outside the World Economic Forum in Davos, Switzerland, on January 23, 2025.
Gerry Miller | CNBC
With Muse Spark, users of the standalone Meta AI app and related website will now be able to alternate between certain modes depending on the sophistication of their prompts. Users can tap one mode to get quick answers to simple questions, and another for more complicated queries related to tasks like analyzing legal documents or gleaning nutritional information from photos of grocery store products.
Additionally, a Contemplating mode “will be rolling out gradually” in the Meta AI app and site for the most complicated queries and tasks, Meta said in the technical blog. For that mode, Muse Spark utilizes 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,” the technical blog said.
The revamped Meta AI with Muse Spark will also contain a Shopping mode that the company said will be able to help people buy clothes or decorate rooms.
“Shopping mode draws from the styling inspiration and brand storytelling already happening across our apps, surfacing ideas from the creators and communities people already follow,” Meta said.
WATCH: Alphabet, Meta, Microsoft all down as data center spending rises.
