According to @AIatMeta, SAM 3 reached roughly 2x the performance of baseline models by leveraging a high quality dataset with 4M unique phrases and 52M corresponding object masks, with the team crediting a data engine for the improvement; source: AI at Meta. According to @AIatMeta, the organization also shared the SAM 3 research paper and emphasized that data scale and quality were central to the performance gains; source: AI at Meta. According to @AIatMeta, the verified catalyst for traders is the performance disclosure itself, so AI-focused participants can monitor flows, liquidity, and volatility in AI narrative assets around the announcement window while awaiting confirmed downstream adoption signals; source: AI at Meta.