SAM 2
Meta's unified segmentation model — click anything in an image or video and it tracks it.
About
SAM 2, released by Meta on July 29, 2024, extends the original Segment Anything Model from images to video. It does real-time, promptable object segmentation — click, box, or mask an object once and the model tracks it across frames, even through occlusion. Meta reports it is six times faster than SAM on images and state-of-the-art on video segmentation. Code and weights are Apache 2.0, plus the 51,000-video SA-V training dataset.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Meta-reported: 6x faster than SAM; SOTA on video object segmentation benchmarks at release.
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 9/10
Scored 9/10 (verified tier, metrics-v1). Evidence: Released July 29, 2024 per Meta; Apache-2.0 weights plus the 51k-video SA-V dataset; Widely adopted in research and products. License: Apache-2.0..
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 9.0/10
- Maintenance · recency, releases, issue load
- 9.0/10
- Openness · license + docs
- 10.0/10
- Manual review · sanity read + red-flag check
- 8.0/10
- · Released July 29, 2024 per Meta
- · Apache-2.0 weights plus the 51k-video SA-V dataset
- · Widely adopted in research and products