Meta's Watermelon AI Matches GPT-5.5 Benchmarks

Meta claims its new AI model, codenamed Watermelon, has matched OpenAI's GPT-5.5 on key performance metrics. This announcement came from Alexandr Wang, Meta’s chief of Superintelligence, during an internal town hall. However, the Watermelon model is still in training. The benchmarks used for comparison are internal and unnamed, and no independent verification has occurred. The model reportedly uses ten times more computational resources than its predecessor, Avocado, which was part of Meta’s Muse Spark release. The specific benchmarks matching GPT-5.5 are not identified. No release timeline has been announced, and the model is not yet finished training. Investors should look for public benchmark releases and a timeline for public access. This information is important for understanding the true progress in AI development.