Skip to content

Griffin Labs

Blog

Research releases and engineering notes: the models we train, the benchmarks we build to test them, and the robots they run on.

11 September 2026
Griffin Alpha: a robust and extensible robot foundation model
A vision-language-action model built for deployment — one set of weights across embodiments, stable over long horizons, and robust when the room changes. Its LIBERO, RoboTwin 2.0 and CleanBench results, released as Griffin Alpha-S.