According to The Verge, Meta is working on a new AI model, internally code-named Avocado, that it might charge companies for access to, as reported by Bloomberg. This marks a potential strategic shift from CEO Mark Zuckerberg, who just last year called “open source AI is the path forward.” The news follows the disappointing launch of the Llama 4 model, after which Zuckerberg scrapped a planned “Behemoth” version to pursue something new. Following that, he has made sweeping changes to the AI team, including hiring former Scale AI CEO Alexandr Wang after Meta invested a staggering $14.3 billion in that company. In a memo on July 30th, Zuckerberg hinted at this shift, stating that to mitigate safety risks, Meta will have to be “careful about what we choose to open source.” He now spends much of his time working closely with new hires in a siloed group called TBD Lab, as detailed in a New York Times report.
The Open Source Facade Cracks
Here’s the thing: Meta’s “open source” stance was always a bit of a marketing flex. The Open Source Initiative itself said Llama’s license wasn’t truly open source. So this move to a paid model feels like dropping the pretense. The Llama 4 launch was a mess, with benchmark gaming and delays. It seems like Zuckerberg looked at the billions being poured into Meta Superintelligence Labs and thought, “We can’t just give this away.” And he’s probably right, from a business perspective. But it’s a huge philosophical reversal. They’re basically admitting that the most advanced, competitive AI might be too valuable—or too dangerous—to release into the wild.
Follow The Money And The Talent
That $14.3 billion investment in Scale AI isn’t just a number; it’s a signal. Meta is betting the company on AI, and you don’t spend that kind of cash to build something you give away for free. Hiring Alexandr Wang, a top-tier AI infrastructure CEO, and creating the exclusive “TBD Lab” bunker near Zuck’s office shows where the real priority lies. This isn’t about democratizing AI for developers anymore. It’s about building a proprietary, market-leading product. They’re assembling a super-team, and super-teams need super-funding. A subscription or licensing model for “Avocado” could be how they start to justify that insane R&D burn rate.
What This Means For The AI Race
So what happens if Meta goes paid? It fundamentally changes the landscape. The open-source community that rallied around Llama would lose its biggest sugar daddy. For businesses, it means one more major vendor to pay, potentially locking them into another ecosystem. But look, it also creates a clearer market segmentation. Free, “open” models for experimentation and lower-tier use, and closed, paid, state-of-the-art models for enterprise applications. Meta is basically conceding that the very top of the AI stack is a commercial battlefield, not a communal playground. It’s a pragmatic, if cynical, move. The question is, will developers and companies who bought into the “open” vision feel burned?
