Meta's Andromeda ads retrieval system scores over one billion ads per second
Overview
Meta says its next-generation Andromeda ads retrieval system, co-designed with NVIDIA GPU hardware, now scores every eligible ad with near ranking-level model performance and surfaces top candidates at over one billion ads per second.
Meta describes this as a 100x throughput lift over its 2024 Andromeda system.
The new architecture uses LLM-inspired mixture-of-experts and runs at roughly 300x fewer inference-time FLOPs than the 2024 Andromeda retrieval model, according to Meta's engineering post. Meta also says Andromeda has delivered a 2.4% increase in Facebook conversions since 2025. The company's report is first-party, and the post does not provide an independent measurement of these figures.
Written by AI from the articles below · updated Oct 28, 12:02 PM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- Engineering at MetaOfficialMeta Andromeda retrieval model scores over one billion ads per second
AIMeta's Andromeda ads retrieval system now scores every eligible ad with near ranking-level model performance and surfaces top candidates at over one billion ads per second, a 100x lift over the 2024 system. The new architecture uses LLM-inspired mixture-of-experts and runs at approximately 300x fewer inference-time FLOPs than the 2024 Andromeda retrieval model. Meta says Andromeda has delivered a 2.4% increase in Facebook conversions since 2025.
Heat Trend
Not enough continuous observations to show a trend yet.