Meta says it has finally pulled level with OpenAI at the frontier of artificial intelligence.
According to Business Insider, Meta's superintelligence chief Alexandr Wang told employees that an upcoming model, codenamed "Watermelon," now matches OpenAI's flagship GPT-5.5 on key AI benchmarks. Benchmarks are standardized tests used across the industry to compare how well rival models reason, code and answer questions.
The model is still in training and has not been released, so these are claims made internally rather than independently verified results. Wang framed the milestone as evidence that Meta is "finally catching up" to OpenAI, per Business Insider.
Techmeme, citing Business Insider's sources, reports that Watermelon uses an "order of magnitude more compute than Avocado" — an apparent reference to an earlier Meta model — signaling a major escalation in the computing horsepower Meta is pouring into the effort.
WIRED characterized the development as giving Meta chief executive Mark Zuckerberg "a seat at the big kid's table," suggesting the company is being taken more seriously among the handful of labs building the most capable systems. Coverage of Wang's remarks also appeared in outlets including Benzinga and GuruFocus.
Meta has spent heavily to compete with OpenAI, Google and Anthropic, and Wang was installed to lead its push toward more powerful AI. A model that genuinely matches GPT-5.5 would mark a turnaround for a company whose earlier releases were seen as trailing rivals.
Why it matters: if Meta's claims hold up once Watermelon ships, the race for the most advanced AI would tighten from a two- or three-horse contest into a broader fight, giving businesses and consumers more competing options for the technology increasingly woven into everyday software.