Meta eyes rivals' AI for its own apps

Meta weighs temporary partnerships with Google and OpenAI while racing to advance its in-house models.

Meta is in talks with Google and OpenAI to use their models in Meta AI and other applications.,These partnerships would be stopgaps until Meta’s own Llama 5 model matures.,Meta Superintelligence Labs is already tapping Anthropic models internally while expanding its AI talent base.

A pragmatic shift in Meta’s AI playbook

The focus keyphrase here is Meta AI partnerships. For years, Meta positioned itself as the open-source counterweight to closed AI ecosystems. Now, a quiet change in tone is emerging. According to reports from The Information, executives inside the company are weighing deals with Google and OpenAI to embed their models directly into Meta’s flagship chatbot and other services.

It is not a retreat but rather a tactical pause. Meta’s in-house models, while popular among researchers, still lag behind rivals in sophistication and scale. Integrating Google’s Gemini or OpenAI’s GPT family into Meta AI would give the company sharper conversational capabilities in the short term, while work continues on its next-generation system, Llama 5.

Meta Superintelligence Labs takes shape

At the centre of this recalibration is Meta Superintelligence Labs, the company’s newly formed AI unit. Led by a mix of star hires and seasoned technologists, the lab is intended to give Meta an edge in the global race to build models powerful enough to match OpenAI’s most advanced offerings.

Earlier this year, Meta committed billions of dollars to lure talent such as Alexandr Wang, the former Scale AI chief executive, and Nat Friedman, once at the helm of GitHub. Both have been tasked with accelerating Meta’s model development while ensuring that the company’s AI infrastructure can scale.

The lab is not afraid to borrow strength where needed. Reports suggest that Meta staff already use Anthropic models for coding assistance inside the company. This hybrid approach reflects what one Meta spokesperson described as “an all-of-the-above strategy”: building, buying, and open-sourcing AI in parallel.

Why partnerships now?

The decision to potentially lean on Google or OpenAI underscores two pressures. First, Meta needs to keep its apps, including Facebook, Instagram, and WhatsApp, competitive in the near term. Conversational AI has rapidly become a consumer expectation, and rivals are embedding increasingly sophisticated assistants across their products.

Second, the development of frontier models is punishingly expensive. Training runs can run into the hundreds of millions of dollars, and access to advanced chips is limited. A stopgap partnership allows Meta to deliver immediate value to its billions of users while buying time for its own research teams to catch up.

Temporary or not, these partnerships signal a willingness to play the long game. Meta knows that its reputation as an open-source champion depends on Llama 5 proving competitive, yet the company cannot afford to let its social apps stagnate while rivals race ahead.

The Asian context

For Asia-Pacific markets, where Meta’s platforms command hundreds of millions of users, the implications are significant. Indonesia, India, and the Philippines rank among Facebook’s largest user bases. Adding advanced AI into these products would not just be a novelty but could change how people search, shop, and communicate online.

Imagine WhatsApp in India integrating a GPT-powered assistant for small businesses, or Instagram in Indonesia offering AI-driven commerce recommendations. In markets where mobile is the first and often only gateway to the internet, these enhancements could quickly become mainstream behaviours.

The question is whether Meta will localise these AI features effectively. Asian languages and cultural contexts are not always well served by models trained predominantly on English and Western data. If Meta chooses to depend on Google or OpenAI models, it risks importing those same blind spots — unless it invests in region-specific fine-tuning. For more on how AI is impacting the region, read our article on APAC AI in 2026: 4 Trends You Need To Know.

Can Llama 5 deliver?

Everything now hangs on Llama 5. Unlike its predecessors, which were open-sourced to the research community, Meta must decide how much of Llama 5 it will share with the outside world. A fully open release would strengthen Meta’s narrative as the champion of democratised AI. A restricted one would signal a tilt towards the closed, commercial model favoured by OpenAI and Google.

Either way, the stakes are high. If Llama 5 arrives as competitive with GPT-5 or Gemini, Meta can reclaim the narrative. If not, it may find itself dependent on the very rivals it hoped to outmanoeuvre. The race to develop advanced AI models is accelerating globally, with countries like South Korea Ramping Into AI Supremacy.

Meta’s balancing act reflects a broader truth about AI in 2025: no company can afford purity. Even giants must blend self-reliance with partnership. The need for running out of data for training models further exacerbates this challenge. For Asia’s millions of Meta users, the real question is whether these moves will deliver AI experiences that feel useful, trustworthy, and local.

Would you trust a Meta AI powered by Google or OpenAI more than one built entirely in-house? For a deeper dive into AI's impact on business, consider reviewing research on the economic impact of AI.

YOUR TAKE

We cover the story. You tell us what it means on the ground.