official announcement

OpenVLA publishes a vision-language-action research model

The original OpenVLA preprint introduced a 7B robot policy trained on 970,000 demonstrations with open adaptation tooling.