SAM 3 runs on-device with LiteRT, supporting text-prompted segmentation
Scobleizer · x · 2026-08-22
SAM 3 (Meta) is now fully runnable on-device via LiteRT for Android, iOS, and macOS. It supports open-vocabulary object detection and instance segmentation via text prompts (e.g., typing "car" segments all cars). The 830M parameter model was tested on Pixel 8a and iPhone 17 Pro, achieving fast re-prompting (as low as 1.3s) by caching vision features, with mask IoU ≥ 0.98 compared to PyTorch.
More from Infra
- TSMC earnings note cross-referenced with WFE model and power checks in CS Atlas — BenBajarin · 2026-10-16
- Fluidstack hires Electrical Engineer for Modular R&D to build frontier AI infrastructure — MxMnr · 2026-10-16
- Jensen Huang joins BG2 podcast: does AI revenue justify capex, '27 bottlenecks — elonmusk · 2026-10-16
- Mistral's on-device infra rack revealed: DGX Spark, Mac Studio and more — gregmushen · 2026-10-16
- NVIDIA and Sweden's Wallenberg launch Sferical One, a shared AI supercomputer — nvidia · 2026-10-16
- Next infra frontier: self-evolving systems that adapt and self-verify — _shreya_s · 2026-10-16