Meta is making great strides in the AI model race, head of superintelligence Alexandr Wang told employees today.
According to two people familiar with the matter, Wang said inside the government building that Meta’s next AI model (codenamed “Watermelon”) has caught up with OpenAI’s flagship GPT-5.5 model. Wang based his results on closely tracked AI model benchmarks. It is not clear which benchmarks Wang cited.
According to sources, Mr. Wang said at the town hall, “We are currently training the watermelon, which is the next model to follow the avocado.” “Watermelon uses orders of magnitude more compute than Avocado,” he added, referring to Meta’s internal code name for Muse Spark, the first product in the model family released in April.
Meta’s AI ambitions have long revolved around the simple goal of closing the gap with OpenAI, Google, and Anthropic. The company has struggled to convince developers and customers that its model is at the cutting edge of the industry, despite massive investments in chips, data centers and human resources.
If Wang’s assessment is correct, it would be the clearest sign yet that Mehta’s investment and Zuckerberg’s aggressive talent blitz are starting to pay off, even as competition continues to move forward at a rapid pace.
GPT 5.5 is a powerful AI model released by OpenAI in April of this year. OpenAI then debuted its most powerful model to date, GPT 5.6, late last month, but it has not yet released it to the public at the request of the US government.
Mehta declined to comment. OpenAI did not respond to a request for comment.
In April, Meta released the first in a series of models called the Muse Spark. The model performed well in our benchmarks, but it did not match or exceed other labs such as OpenAI and Anthropic.
Zuckerberg is pushing hard for Meta to gain an edge in the AI race. Last year, he appointed Wang to lead the effort and renamed the company’s AI division Meta Superintelligence Labs.
At Meta, Wang oversees a team of elite AI researchers known as TBD and other AI efforts, including recent hardware advances. Business Insider previously reported that Meta is offering hundreds of millions of dollars each to top AI talent to join.
Talent acquisition will accelerate as Meta increases spending on infrastructure. The company told investors this year that it expects to spend $125 billion to $145 billion on chips, data centers and other infrastructure this year, up from previously expected spending of $115 billion to $135 billion, due to higher component costs and additional spending on data centers.
Any tips? Contact Charles by email. crolet@businessinsider.com or on Signal and WhatsApp 628-282-2811. Email Pranav. pdixit@insider.com or Signal at 408-905-9124. Use a personal email address, non-work WiFi network, and non-work device. Here’s a guide to sharing your information securely.