Meta's Watermelon AI model reportedly matches OpenAI's GPT-5.5 on key benchmarks

2 Sources

Share

Meta's superintelligence chief Alexandr Wang told employees the company's upcoming Watermelon AI model matches OpenAI's GPT-5.5 on certain benchmarks. The announcement comes as Mark Zuckerberg invests heavily in AI infrastructure and talent acquisition to compete in the generative AI space, though specific performance metrics remain undisclosed.

Meta AI Model Watermelon Claims Parity with OpenAI's GPT-5.5

Alexandr Wang, Meta's superintelligence chief, announced during an internal town hall meeting that the company's upcoming Meta AI model codenamed Watermelon has matched OpenAI's GPT-5.5 on certain AI model benchmarks, according to

Business Insider

2

. The claim, reported by two sources familiar with the matter, marks a significant milestone for Meta's competitiveness in AI as the company attempts to close the gap with leading rivals in the generative AI space. However, the exact benchmarks Wang referenced have not been disclosed, leaving questions about the scope and nature of this performance comparison

1

.

Source: Benzinga

Source: Benzinga

Wang explained that Watermelon represents Meta's next major AI development step following Avocado, which is internally referred to as Muse Spark. The new model is currently in training and demands significantly more computing power than its predecessor, underscoring the escalating resource requirements in advanced AI development

1

. Meta's upcoming AI model signals the company's determination to compete with established leaders like OpenAI and Anthropic, despite historically lagging behind these competitors.

Mark Zuckerberg's Aggressive AI Infrastructure Investment

The announcement comes amid CEO Mark Zuckerberg's substantial investment in AI infrastructure, including billions spent on chips, data centers, and talent acquisition to strengthen Meta's position in the competitive landscape

2

. Zuckerberg has personally overseen AI development efforts and brought Wang on board last year specifically to lead Meta's AI division. This hands-on approach reflects the strategic priority Meta places on catching up with rivals who have dominated public perception in the AI race.

Meta released its Muse Spark model in April, which performed well on benchmarks but still fell short of rival offerings from OpenAI and Anthropic

2

. Mizuho Securities views Meta's early Muse Spark launch as a positive indicator of the company's AI competitiveness, noting that increased usage, particularly through Shopping mode, could significantly boost monetization via search and ad targeting

1

.

What This Means for the AI Race

The claim that Watermelon matches OpenAI's performance raises important questions about how AI model benchmarks are measured and compared across different companies. Without transparency about which specific metrics were used, it remains difficult to assess whether Meta has truly achieved parity across all dimensions of model capability or only on select tests. Watch for independent verification of these claims and official benchmark disclosures as Watermelon moves closer to release. The computing power requirements Wang mentioned suggest Meta is scaling up its infrastructure investments, which could have implications for the company's capital expenditure and profitability in coming quarters. Meta Platforms did not immediately respond to requests for comments on Wang's announcement

1

.

Today's Top Stories

© 2026 TheOutpost.AI All rights reserved