Meta says SAM 3.1 makes real-time video object tracking cheaper to run
Meta says SAM 3.1 uses object multiplexing to track up to 16 objects per pass, doubling throughput on an H100 GPU and cutting redundant compute.
Quick answer
Instead of re-watching a video separately for each object you care about, the model tracks many objects at once, which saves time and memory.
What happened
On March 27, 2026, Meta introduced SAM 3.1 as a drop-in update to SAM 3, focusing on faster video tracking by processing multiple objects together instead of one pass per object.
Why it matters
Segmentation and tracking models power video editing, robotics, and monitoring. If the same accuracy can be delivered with fewer GPU passes, more teams can run these workflows in real time on smaller hardware.
Key points
- Object multiplexing lets the model track up to 16 objects in a single forward pass.
- Meta reports throughput rising from 16 to 32 frames per second on a single H100 for medium-object videos.
- Positioned as a drop-in replacement for SAM 3 to reduce redundant computation.
What to watch
Watch for independent benchmarks on accuracy vs. speed trade-offs, and whether downstream tools adopt SAM 3.1 checkpoints as the new default for video workflows.
Key terms
- Segmentation
- Labeling which pixels belong to an object in an image or video frame.
- Forward pass
- A single run of a neural network to compute outputs from inputs.
Sources
- SAM 3.1: Faster and More Accessible Real-Time Video Detection and Tracking With Multiplexing and Global Reasoning Meta AI · Primary announcement · Mar 27, 2026 Primary