Meta Releases SAM 3.1 with Object Multiplexing to Double Video Tracking Speed

Mar 28, 2026 · Updated Apr 25, 2026

AI at Meta released SAM 3.1, a drop-in update to its SAM 3 model that optimizes real-time video segmentation. The core innovation is object multiplexing, which enables the model to track up to 16 objects simultaneously in one forward pass. This replaces the previous requirement for a dedicated pass per object.

This architectural shift addresses memory bottlenecks and redundant computations inherent in multi-object tracking. By processing objects together, SAM 3.1 doubles throughput for videos with a medium number of objects, increasing speed from 16 to 32 frames per second on a single H100 GPU. These gains are achieved without sacrificing accuracy.

The update makes high-performance computer vision feasible on smaller, more accessible hardware. Because it is a drop-in replacement for SAM 3, you can adopt the new model checkpoints and codebase immediately. This is relevant for building real-time tracking tools or dense video segmentation pipelines.

View the full update on ai.meta.com

AI at Meta

@AIatMetaMar 27

We’re releasing SAM 3.1: a drop-in update to SAM 3 that introduces object multiplexing to significantly improve video processing efficiency without sacrificing accuracy. We’re sharing this update with the community to help make high-performance applications feasible on smaller, more accessible hardware. 🔗 Model Checkpoint: https://t.co/KgW0zZQ0QT 🔗 Codebase: https://t.co/Ks61vfokB0

1731.5k

View on X