Meta Stock: Alexandr Wang Says Watermelon AI Has Caught Up With OpenAI

TLDR

• Meta’s superintelligence chief Alexandr Wang told employees that the upcoming Watermelon AI model has caught up with OpenAI’s GPT-5.5.
• Watermelon uses “an order of magnitude” more compute than Avocado, the internal name for the recently released Muse Spark model.
• Meta’s stock fell 4.90% following the report, despite the positive internal claims.
• Meta plans to spend between $125–$145 billion on AI infrastructure in 2026, up from an earlier $115–$135 billion forecast.
• Wang teased “major gains” in coding and agentic capabilities in an upcoming Muse Spark update.

Meta (META) stock dropped 4.90% on Friday even as the company’s internal AI progress appears to be accelerating. Superintelligence chief Alexandr Wang told employees at a town hall that Meta’s upcoming model, codenamed Watermelon, has now matched OpenAI’s flagship GPT-5.5 — a claim that, if accurate, would mark a major milestone for the company’s AI push.

Meta Platforms, Inc., META

Wang cited AI model benchmarks to back up the claim, though it’s not clear which specific benchmarks he referenced.

Watermelon is Meta’s next model after Avocado — the internal codename for Muse Spark, which the company released in April. Muse Spark performed well on benchmarks at the time but still fell short of top offerings from OpenAI and Anthropic.

Wang said Watermelon uses “an order of magnitude” more compute than Avocado. More compute typically means more powerful results, though it also means higher costs.

Coding and Agentic Gains on the Way

Wang also went public with some of the progress. In a post on X, he said an update to the current Muse Spark model is coming soon, with major improvements in coding and agentic capabilities.

When asked by a user when Meta would have a coding model on par with Anthropic’s Claude Opus, Wang replied it would be “pretty soon” and told users they’d like what the company has “cooking.”

Meta has spent years trying to close the gap with OpenAI, Google, and Anthropic. Heavy investment in chips, data centers, and top AI talent has not yet translated into market leadership — at least not publicly.

Zuckerberg hired Wang last year to lead what was rebranded as Meta Superintelligence Labs. Wang oversees a team of elite researchers internally known as TBD, along with the company’s broader AI and hardware efforts.

Infrastructure Spending Keeps Climbing

Meta has offered top AI talent hundreds of millions of dollars each to join the team. The spending doesn’t stop there.

The company told investors it now expects to spend between $125 billion and $145 billion in 2026 on chips, data centers, and infrastructure. That’s up from an earlier forecast of $115–$135 billion, driven by rising component costs and expanded data center plans.

For context, OpenAI released GPT-5.5 in April. It then debuted GPT-5.6 late last month — described as its most powerful model yet — though it hasn’t been made broadly available yet, reportedly at the request of the US government.

Wang’s claims have not been independently verified, and Meta declined to comment. OpenAI did not respond to a request for comment.

The post Meta Stock: Alexandr Wang Says Watermelon AI Has Caught Up With OpenAI appeared first on CoinCentral.