Meta Platforms is making significant strides in the artificial intelligence (AI) race, according to CEO Alexandr Wang, who told employees during an internal town hall that Meta’s upcoming AI model, codenamed Watermelon, has caught up with OpenAI’s GPT-5.5.
During the town hall, Wang stated, “Watermelon, our next model after Avocado, is currently in training.” He further noted that Watermelon uses significantly more computational resources compared to Meta’s previous internal model, Muse Spark (codenamed Avocado), which was released in April.
In a post on X (formerly Twitter) on Thursday, Wang alluded to the progress publicly. “An update to our current model, Muse Spark, is coming soon with major gains in coding and agentic capabilities aimed at closing the gap with rival models,” he wrote. When asked when Meta would have a coding model that matches Anthropic’s Claude Opus, Wang replied, “It will be pretty soon,” adding that users would like what the company has “cooking.”
Meta’s AI ambitions have long centered on narrowing the gap with industry leaders OpenAI, Google, and Anthropic. Despite substantial investments in hardware and talent, the company has struggled to gain a leading edge among developers and customers. However, if Wang’s assessment is accurate, it would mark a significant milestone in Meta’s efforts.
Wang oversees the AI division now known as Meta Superintelligence Labs, which has seen a massive talent acquisition push, with top AI researchers offered hundreds of millions of dollars each to join. According to Business Insider, this aggressive talent recruitment comes amid plans for substantial spending on infrastructure. Meta expects to invest between $125 billion and $145 billion in chips, data centers, and other technology, up from an earlier forecast of $115 billion to $135 billion.
The internal codenames Avocado for Muse Spark and Watermelon for the next iteration suggest a methodical progression with each generation leveraging significantly more computational resources. Wang indicated that Meta is focusing on practical improvements in areas like coding and agentic capabilities, which are critical for real world utility.
OpenAI continues to set the pace with its GPT series, while Anthropic’s Claude models have gained traction in enterprise settings. Google’s Gemini family also presents formidable competition, particularly due to its integration with Android and other Google services. For Meta, success in AI isn’t just about matching benchmarks; it involves translating technical progress into products that drive user engagement across its suite of apps.
With Zuckerberg’s commitment to not falling behind in what many see as the defining technology of the era, Meta is pouring resources into attracting top talent and scaling compute power. The company’s efforts could position it more competitively in the AI race, but the journey remains challenging.
Thinking about building an AI product?
Get in Touch