Key Info

Meta's Segment Anything Model (SAM) 3.1 is now available on the Meta Model API, providing developers with a fast, lightweight model that can perform detection, segmentation and tracking in a single call, with inference tuned specifically for SAM 3.1's architecture.

Highlights

  • Use a short text phrase to find objects in both images and video.
  • One API call returns object detections, pixel-precise segmentation masks, and identity-preserving video tracks.
  • The model is positioned as fast and lightweight, aiming to simplify building with unified vision capabilities via the Meta Model API.