Meta's 'Watermelon' AI Model Catches Up to GPT-5.5, Superintelligence Chief Tells Employees

Meta's superintelligence chief Alexandr Wang told employees at an internal town hall that the company's next AI model — codenamed Watermelon — has caught up with OpenAI's GPT-5.5 on closely followed AI benchmarks. Wang didn't specify which benchmarks he cited, but the claim marks a significant milestone for a company that has consistently trailed OpenAI, Google, and Anthropic at the frontier of AI capability.

Watermelon is the successor to an internal model codenamed Avocado — Meta's name for Muse Spark, the first in a family of models released publicly in April. Wang told employees that Watermelon uses an order of magnitude more compute than Avocado, signaling a dramatic scaling push. Muse Spark performed well on benchmarks at launch but didn't match or exceed rival models from OpenAI or Anthropic.

The timing is notable: OpenAI released GPT-5.5 in April, then debuted an even more powerful GPT-5.6 late last month — though that model hasn't been broadly released yet, reportedly at the request of the US government. That means even if Meta has closed the gap with GPT-5.5, it's chasing a target that has already moved.

Wang also hinted publicly at the progress. In a post on X, he said an updated version of Muse Spark is coming soon with major gains in coding and agentic capabilities. Asked when Meta would have a coding model competitive with Anthropic's Claude Opus, Wang replied it would be 'pretty soon.'

The developments reflect Mark Zuckerberg's aggressive push to win the AI race. He appointed Wang — founder of Scale AI — last year to lead the effort, rebranding the division as Meta Superintelligence Labs. Meta has been recruiting elite AI researchers with compensation packages reportedly reaching hundreds of millions of dollars per person, and has raised its 2025 infrastructure spending forecast to between $125 billion and $145 billion, up from an earlier estimate of $115 billion to $135 billion, citing rising component costs and expanded data center buildout.