Meta believes that its new neural network is on par with OpenAI's flagship model, BI

Meta’s next artificial intelligence model, codenamed Watermelon, has caught up with OpenAI’s flagship GPT-5.5. Alexander Wang, head of Meta’s AI division, made this announcement at an internal company meeting, according to Business Insider, which cites two sources. GPT-5.5 is a state-of-the-art neural network that OpenAI released in April of this year.

“Watermelon, our next model after Avocado, is currently in the training phase,” Wang said. “It requires an order of magnitude more computing power than its predecessor.” By “Avocado,” the top executive was referring to the internal codename for the Muse Spark neural network, which Meta unveiled in April.

Business Insider points out that Wang backed up his claims with references to authoritative benchmarks, though it is unclear exactly which tests he was referring to.

The head of Meta’s AI division also publicly hinted at the company’s progress. On July 2, he posted on the social media platform X that an update to the current version of Muse Spark would be released soon. The release will include significant improvements in code generation and the performance of AI agents, which should help close the gap with competitors. When asked by a user when Meta would release a programming model on par with Anthropic’s Claude Opus, Wang replied, “Pretty soon,” adding that users would definitely like what the company is currently “working on.”

Meta declined to comment to Business Insider. OpenAI did not respond to the request.

Meta CEO Mark Zuckerberg aims to position the company as a leader in the AI industry, according to Business Insider. As part of this strategy, Alexander Wang was appointed head of the AI division last year. This year alone, Meta plans to invest up to $145 billion in the development of its AI infrastructure.

For a long time, Meta’s efforts were focused on narrowing the technological gap with OpenAI, Google, and Anthropic. However, the first model in the Muse Spark family, unveiled in April, failed to compete with developments from other labs, despite its strong test results. As a result, despite massive investments in microchip procurement, data center construction, and hiring specialists, the company has failed to convince the market of the competitiveness of its solutions, according to Business Insider.

Against this backdrop, the potential success of the Watermelon project is crucial, the publication explains. If Van’s statement is confirmed, it will be the most compelling evidence that Meta’s capital investments and Zuckerberg’s strict hiring policies are beginning to pay off, despite the industry’s rapid growth.

Meta’s development of AI agents over the past four months has not met management’s initial expectations, Zuckerberg said during an internal staff meeting (Reuters hasa recording of the speech). The top executive also acknowledged that the company’s reorganization and large-scale layoffs had been difficult, and that the new structure has not yet yielded results. Zuckerberg expects to see a more significant return on investment in AI within the next 3–6 months.

This article was AI-translated and verified by a human editor