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 isn’t positioning Muse Spark as a top-of-the-line model, but is instead highlighting its efficiency and “competitive performance” on various tasks.

The new Muse Spark will be proprietary, instead of open source, 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.”

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 that 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.

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 get use one mode 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.

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. With Instant mode, users can get quick answers to simple questions whereas Thinking mode lets them input 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. In this mode, the Muse Spark model utilizes a squad of AI agents to help “reason in parallel,” thus 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’s blog post said.

This is breaking news. Please check back for updates.

WATCH: Alphabet, Meta, Microsoft all down as data center spending rises.

As AI chipmaker Cerebras angles for an eventual IPO, the company appears to have landed a significant cloud-computing customer: Oracle.

On a conference call with analysts on Tuesday following Oracle’s quarterly earnings, Clay Magouyrk, one of the software vendor’s two CEOs, indicated that his company’s infrastructure includes Cerebras chips, alongside graphics processing units (GPUs) from market leader Nvidia and rival Advanced Micro Devices.

“We build infrastructure which is flexible, fungible, and can support the smallest workloads up to the largest,” Magouyrk said. “We continually offer the latest in accelerators, from the most recent Nvidia and AMD options to emerging designs from companies like Cerebras and Positron,” another AI hardware startup.

Cerebras offers cloud services that employ its large-scale WSE-3 chips. The company filed paperwork for an IPO in 2024 but withdrew the filing last October. Days later, it announced a $1.1 billion funding round at a valuation of $8.1 billion, and CEO Andrew Feldman said Cerebras still intends to go public.

For prospective investors, one of the most glaring concerns from Cerebras’ original prospectus was its reliance on a single customer based in the Middle East. G42, backed by Microsoft, is headquartered in Abu Dhabi, United Arab Emirates, and in the first half of 2024, it accounted for 87% of Cerebras’ revenue.

Bolstering its client roster with a name like Oracle could be a big boon for Cerebras, and it would follow another significant announcement earlier this year. In January, Cerebras said it had received a $10 billion commitment from OpenAI, which relies on Oracle, and other companies, for cloud services. The next month, OpenAI said it was collaborating with Cerebras on a research preview of Codex-Spark, a fast-acting AI model geared toward software development, for ChatGPT Pro customers.

Oracle didn’t immediately respond to a request for comment, and its price list does not mention a Cerebras option. Cerebras didn’t immediately provide a comment.

Oracle’s earnings call came after the company reported better-than-expected results, lifted its fiscal 2027 guidance and said remaining performance obligations more than quadrupled to $553 billion from a year earlier.

“Altogether, we are confident that the investments we make now in data centers, compute capacity and customer relationships will only grow more valuable over time,” Magouyrk said, after naming Cerebras and other chipmakers.

While Cerebras is trying to compete as an upstart against the world’s most valuable company, it’s playing in a market with seemingly insatiable demand for computing power as AI model developers scale to quickly respond to the needs of users.

Nvidia is using its mammoth cash pile to expand into new product areas. In December, the company bought key assets from AI chip startup Groq for about $20 billion. Nvidia plans to announce a new architecture drawing on Groq at its GTC developer conference in California next week, The Wall Street Journal reported.

Magouyrk said on the call that GTC will feature some “key announcements.” He also said that speed in responding to incoming requests requires innovative technology in addition to strategically located data centers.

“It’s the type of hardware that’s being deployed, and that’s why you’re seeing so much innovation going on around these AI accelerators,” he said. “If you look at what Groq does, or Cerebras or Positron, all of these different types of customers are saying, well, not only how do we reduce the cost of inferencing, but also, how can we significantly reduce the latency of it?”

WATCH: OpenAI unveils first AI model running on Cerebras chips

Oracle shares rose as much as 10% in extended trading on Tuesday after the software vendor reported quarterly results that surpassed Wall Street projections and boosted its revenue guidance for fiscal 2027.

Oracle sees $1.92 and $1.96 in adjusted earnings per share for the fiscal fourth quarter, with revenue growth between 19% and 20%. LSEG’s consensus included $1.70 per share and 20% revenue growth.

Here’s how the company did in the quarter relative to LSEG consensus:

• Earnings per share: $1.79 adjusted vs. $1.70 expected
• Revenue: $17.19 billion vs. $16.91 billion expected

Oracle’s overall revenue increased 22% year over year in the fiscal third quarter, which ended on Feb. 28, according to a statement. Net income rose to $3.72 billion, or $1.27 a share, from $2.94 billion, or $1.02 a share, in the same quarter a year earlier. Adjusted earnings per share excludes stock-based compensation expense.

The company reported $8.9 billion in total cloud revenue, including infrastructure and software as a service, or SaaS. The number was up 44% and more than the $8.85 billion consensus among analysts surveyed by StreetAccount.

Management pushed up the company’s fiscal 2027 revenue forecast by $1 billion to $90 billion. Analysts polled by LSEG had anticipated $86.6 billion.

Oracle said it generated $4.9 billion in cloud infrastructure revenue, up 84%, a faster pace than the 68% growth in the prior quarter. The company touted cloud business from Air France-KLM, Lockheed Martin, SoftBank Corp. and Microsoft’s Activision Blizzard video game subsidiary.

Shares of Oracle have plummeted over 50% from their September highs, falling along with other software vendors on broader artificial intelligence concerns as well as Wall Street’s specific fears about the company’s hefty debt load that’s funding its AI buildout.

Thank God we have these coding tools now that allow us to build a comprehensive set of software, agent-based software, to implement, to automate a complete ecosystem like healthcare or financial services,” Larry Ellison, Oracle’s co-founder, technology chief and executive chairman, said on a conference call with analysts. “That’s what we’re doing at Oracle. That’s why we think we’re a disruptor. That’s why we think the SaaS apocalypse applies to others but not to us.”

As of Tuesday’s close, the stock had declined 23% in 2026, while the S&P 500 is down less than 1% in the same period.

Oracle has won large contracts to deliver cloud infrastructure to AI companies such as OpenAI, but has less cash on hand than larger competitors such as Amazon and Microsoft.

Renting out Nvidia graphics chips ekes out a smaller profit margin than selling software licenses, and Oracle reported $13.18 billion in negative free cash flow for the past 12 months.

During the quarter, Oracle announced plans to raise $45 billion to $50 billion in the fiscal year to expand its cloud infrastructure capacity. The company is planning for over 10 gigawatts worth of computing power coming online in the next three years, Clay Magouyrk, its other CEO, said on the call.

The across-the-board beat may help settle a nervous investor base, at least for the time being, as Oracle’s results and backlog point to a continuing surge in demand for AI infrastructure. Remaining performance obligations more than quadrupled to $553 billion from a year earlier — although it was slightly lower than StreetAccount’s $556 billion consensus — and the company said it has the capital to support that growth.

“Most of the increase in RPO in Q3 related to large scale AI contracts where Oracle does not expect to have to raise any incremental funds to support these contracts as most of the equipment needed is either funded upfront via customer prepayments so Oracle can purchase the GPUs, or the customer buys the GPUs and supplies them to Oracle,” the company said in the statement.

In Abilene, Texas, where Oracle and Crusoe are constructing a data center project for OpenAI, “two buildings are completely operational and the rest of the campus is on track,” Oracle said in a Sunday X post. The statement came after Bloomberg reported that Oracle and OpenAI had dropped plans to expand the site, though Oracle said media reports regarding Abilene were incorrect.

At the end of February, Oracle announced a $110 funding round, with backing from Amazon and Nvidia, among others.

“Some of the largest consumers of AI Cloud capacity have recently strengthened their financial positions quite substantially,” Oracle said in its Tuesday statement.

Bloomberg reported last week that Oracle was planning layoffs.

“AI models for generating computer code have become so efficient that we have been restructuring our product development teams into smaller, more agile and productive groups,” Oracle said in the statement. “This new AI Code Generation technology is enabling us to build more software in less time with fewer people. Oracle is now building more SaaS applications for more industries at a lower cost.”

— CNBC’s Ari Levy contributed to this report.

WATCH: Inside Oracle’s risky AI bet