Meta is accelerating its push to lead the artificial intelligence landscape, with AI division chief Alexandr Wang announcing that the company’s forthcoming Watermelon model has reached parity with OpenAI’s flagship GPT-5.5. During an internal town hall, Wang confirmed that Watermelon, which follows the April release of the Muse Spark series codenamed Avocado, is currently in training and utilizes an order of magnitude more computational resources than its predecessor. The claim is reportedly based on widely tracked industry benchmarks, marking a notable milestone in Meta’s multi-year effort to close the performance gap with OpenAI, Google, and Anthropic. This development underscores the early returns of CEO Mark Zuckerberg’s aggressive strategy to overhaul Meta’s artificial intelligence capabilities. Since appointing Wang to lead the newly rebranded Meta Superintelligence Labs, the company has executed a massive capital expenditure campaign, increasing its annual infrastructure forecast to between 125 billion and 145 billion dollars to fund advanced chips, data centers, and software development. Simultaneously, Meta has launched a high-profile talent acquisition drive, reportedly offering elite researchers substantial financial packages to join its newly formed TBD research team and accelerate model development. While Meta’s previous April release demonstrated strong benchmark performance, it ultimately fell short of rival offerings. Watermelon’s reported alignment with GPT-5.5 suggests a significant leap in capability. OpenAI recently debuted GPT-5.5 in April and is currently withholding its more advanced GPT-5.6 model following requests from the US government. Meta declined to comment on the internal assessment, and OpenAI did not respond to inquiries. If validated, Wang’s announcement would represent the clearest indicator yet that Meta’s sustained investments in hardware, data, and human capital are successfully positioning its next-generation models at the industry’s technological frontier.