Meta Watermelon AI Model Catches GPT-5.5, Alexandr Wang Tells Staff

Meta has caught up with OpenAI’s GPT-5.5 in closely watched AI benchmarks, superintelligence chief Alexandr Wang told employees.

Meta Watermelon

Wang made the comments during an internal town hall, according to two people familiar with the matter.

He said Meta’s next model, codenamed Watermelon, had caught up with OpenAI’s flagship GPT-5.5 model based on closely watched AI benchmarks. The people did not say which benchmarks he cited.

“Watermelon, our next model after Avocado, is currently in training,” Wang said, according to one person familiar with the meeting.

“Watermelon uses an order of magnitude more compute than Avocado,” he added, referring to Meta’s internal codename for Muse Spark, the first model in a family released in Apr.

Wang also pointed to progress publicly in a post on X on Thursday. He said an update to Muse Spark would arrive soon with stronger coding and agentic abilities. Asked when Meta would have a coding model on par with Anthropic’s Claude Opus, Wang replied that it would be “pretty soon.” He added that users would like what the company had “cooking.”

Meta declined to comment. OpenAI did not respond to a request for comment.

Also Read: Hollywood Director Gets 30 Months After Crypto And Stock Bets

Wang AI Push

Wang’s comments matter because Meta has spent heavily to close the gap with OpenAI, Google and Anthropic.

The company has invested in chips, data centers and AI talent, but it has not fully convinced developers and customers that its models sit at the industry’s leading edge.

If Wang’s assessment is accurate, Watermelon would be the clearest sign that Mark Zuckerberg’s AI strategy is beginning to produce results. Zuckerberg appointed Wang last year to lead the effort and renamed the AI division Meta Superintelligence Labs.

Wang now oversees elite AI researchers known as TBD, along with other AI work, including a recent hardware push.

Meta has offered top AI researchers hundreds of millions of dollars each to join, Business Insider previously reported.

That recruitment drive comes as Meta raises infrastructure spending. The company told investors this year that it expects to spend $125 billion to $145 billion on chips, data centers and other infrastructure, up from its earlier $115 billion to $135 billion forecast.

Meta’s last major model step came in Apr., when it released Muse Spark. The model performed well on benchmarks but did not match or surpass OpenAI or Anthropic, leaving Watermelon as the next major test of Meta’s frontier AI ambitions.

Read Next: Record 207 Crypto Hacks Expose Weak Spot In Digital Asset Security