# Berkeley Lab Runs Meta's Open Vision Models Inside Government Compute for the Genesis Mission

Running Segment Anything 3 and DINOv3 on 300 A100 accelerators at the National Energy Research Scientific Computing Center cut one beamline annotation step from about a month to roughly 15 minutes.

- Published: 2026-09-27T06:30:19.445Z
- Canonical: https://polylog.news/ai/2026-09-27/berkeley-lab-runs-meta-s-open-vision-models-inside-governmen
- Publisher: Polylog (AI desk)
- Section: tech
- Sources: [Meta AI](https://ai.meta.com/blog/genesis-mission-lawrence-berkeley-national-laboratory-segment-anything-dino/), [Google Research](https://research.google/research-areas/)

Lawrence Berkeley National Laboratory is running two of Meta's open-weight vision models, Segment Anything 3 and DINOv3, on 300 NVIDIA A100 accelerators at the National Energy Research Scientific Computing Center, as part of the Department…

This story is for subscribers. Read it in full at https://polylog.news/ai/2026-09-27/berkeley-lab-runs-meta-s-open-vision-models-inside-governmen (subscription information: https://polylog.news/pricing).