🟡 📦 Open Source Published: · 2 min read ·

Meta: SAM 3 and DINOv3 open-sourced for the Genesis Mission, segmentation analysis cut from a month to 15 minutes

Scientist analyzing a micro-CT scan of a grapevine stem with AI image segmentation

Meta has open-sourced the SAM 3 and DINOv3 image segmentation models as part of the SYNAPS-I project, part of the White House Genesis Mission initiative for AI in science. The US Department of Energy generates tens of petabytes of data annually; segmentation that used to take a month now takes about 15 minutes using 300 NVIDIA A100 GPUs and 60 researchers from five national laboratories.

🤖

This article was generated using artificial intelligence from primary sources.

What is Meta open-sourcing, and for which project?

Meta has open-sourced SAM 3 (Segment Anything Model 3) and DINOv3, two image-analysis models, as part of the SYNAPS-I project — part of the White House Genesis Mission initiative, launched in late 2025 to accelerate the application of artificial intelligence in scientific research. SAM 3 is a model for image segmentation, the automatic extraction and delineation of objects within visual data, while DINOv3 is a model for recognizing visual features without the need for supervised (human-labeled) training.

From a month of expert annotation to 15 minutes

Facilities of the US Department of Energy (DOE) generate tens of petabytes of visual data annually, with capture rates of 100,000 images per second on some instruments. Segmentation analysis of such data, which previously required about a month of expert manual annotation, is now completed in roughly 15 minutes thanks to the SAM 3 and DINOv3 models running on 300 NVIDIA A100 GPUs at NERSC and other national supercomputers. The difference between a month of human labor and 15 minutes of automated processing shows the scale of acceleration open models bring to scientific teams facing enormous data volumes.

Five national laboratories and concrete proof on grapevines

The project involves 60 researchers from five national laboratories — Lawrence Berkeley in the lead role, alongside Argonne, Brookhaven, and Oak Ridge. DOE Under Secretary Dario Gil gave a statement supporting the initiative. As a concrete example of application, Meta and DOE demonstrated using the models in research on grapevine drought resistance, where micro-CT scans of grapevine stems are now segmented automatically instead of by hand.

Open-sourcing SAM 3 and DINOv3 under an open-source license means national laboratories and research institutions outside Meta can adopt both models and apply them to their own datasets without additional licensing, extending the reach of the Genesis Mission initiative beyond the SYNAPS-I pilot project to potentially the entire network of US national laboratories.

Frequently Asked Questions

What are SAM 3 and DINOv3?
SAM 3 (Segment Anything Model 3) is Meta's open-source model for image segmentation — automatically extracting objects from visual data — while DINOv3 is a model for visual feature recognition without supervised training; both are now available as open source for scientific use.
What is the Genesis Mission?
The Genesis Mission is a White House initiative launched in late 2025 to accelerate the application of artificial intelligence in scientific research within US national laboratories.
How much faster is segmentation analysis of scientific data now?
Segmentation analysis that previously required about a month of expert manual annotation is now completed in roughly 15 minutes with the help of 300 NVIDIA A100 GPUs at NERSC and other national supercomputers.

📬 AI news in your inbox

A daily digest built your way — pick topics, sources and cadence. One-click unsubscribe.