Meta’s Generative Ads Model (GEM): Scaling the Central Model and Compute Engine for Frontier Ads AI
Meta's unified GEM foundation model delivers 18.9% conversion and 12% CTR improvement across all ad surfaces
“18.9% conversion improvement and 12% ad click-through rate (CTR) improvement”
Meta's 2026 GEM consolidates previously siloed per-domain ad models (Facebook, Instagram, Reels) into a single LLM-scale foundation model using a new architecture called Kunlun, co-designed with the latest GPU hardware. The system achieves LLM-like scaling laws in recommendation systems—a historically difficult problem—enabling 2.5x model complexity growth and meaningful business gains across the full ad funnel. This is a significant signal that LLM-style unified foundation model paradigms are now displacing fragmented task-specific ML systems even in high-stakes, latency-sensitive production ad infrastructure.