Meta’s upcoming AI model, codenamed Watermelon, reportedly utilizes significantly more computing resources than its predecessor, Avocado, according to statements made by Alexandr Wang during an internal meeting. Wang noted that Watermelon has positioned itself to rival OpenAI’s GPT-5.5, reflecting Meta’s ongoing efforts to enhance its AI capabilities in a highly competitive landscape where industry giants are vying to advance large language model technology.

Meta: Meta is a technology company that develops social media platforms and invests heavily in artificial intelligence research and infrastructure. The company is currently preparing to release an upcoming AI model internally codenamed Watermelon. According to reports, this model has reached performance levels comparable to OpenAI’s GPT-5.5.
Watermelon: Watermelon is the internal codename for Meta’s next-generation AI model under development. It is described as utilizing substantially greater computational resources than the company’s prior model, referred to as Avocado. The model is positioned as having caught up to the capabilities of OpenAI’s GPT-5.5.
Alexandr Wang: Alexandr Wang is an AI industry executive who has commented publicly on developments at major labs. In this case, he referenced details from an internal Meta meeting regarding the compute requirements of the company’s Watermelon model relative to its predecessor.

{“AI Development”: “Major technology firms continue to develop advanced language models under various codenames to enhance their AI capabilities.”, “Industry Competition”: “Meta is focused on diminishing the performance discrepancies with AI systems created by its competitors, such as OpenAI.”}