OpenVLA
lastmod 2026-11-10created 2026-01-152 edits
OpenVLA is an open-source Vision-Language-Action model released for robot manipulation. It combines a pretrained vision encoder with a large language model and then fine-tunes them on a large-scale manipulation dataset collected from multiple robot platforms, taking camera images and natural language instructions as input and outputting robot actions as tokens. Its model weights and training code are public, so it can be fine-tuned efficiently for new robots and tasks.
Documentation
- OpenVLA - An Open-Source Vision-Language-Action Model
- https://arxiv.org/abs/2406.09246