# VLA Robotics Benchmark Index (vlabench.dev)

> Primary-source index of vision-language-action models, manipulation benchmarks, and robot datasets

## Canonical Definition
A vision-language-action model maps camera images and a language instruction to robot motor commands. This index records public VLA models, manipulation benchmarks, and training datasets, copying figures only from author papers and project pages retrieved on 19 October 2026. Records include OpenVLA, RT-2, π0, Octo, LIBERO, and Open X-Embodiment.

## Machine-Readable Data Endpoints
- [Primary JSON Dataset](https://vlabench.dev/data/dataset.json)
- [Primary CSV Dataset](https://vlabench.dev/data/dataset.csv)
- [ARD v1.0 Discovery Manifest](https://vlabench.dev/.well-known/ard.json)
- [Verification Methodology](https://vlabench.dev/methodology/)

## Verified Entities (14)
- [OpenVLA](https://vlabench.dev/entity/openvla/): 7B open VLA trained on 970k robot demonstrations, using Llama 2 with fused DINOv2 and SigLIP features. (Status: arXiv v3, Metric: 7B; 970k demos)
- [OpenVLA-OFT](https://vlabench.dev/entity/openvla-oft/): Stanford fine-tune of OpenVLA with 97.1% average success on four LIBERO suites and 26x faster actions. (Status: Project release, Metric: 97.1% LIBERO avg)
- [RT-2](https://vlabench.dev/entity/rt-2/): Google DeepMind VLA that writes robot actions as text tokens, at up to 55B parameters over 6k trials. (Status: arXiv 2023, Metric: Up to 55B; 6k trials)
- [RT-1](https://vlabench.dev/entity/rt-1/): Robotics at Google transformer reporting 97% success on 700 instructions from 130k episodes on 13 robots. (Status: RSS 2023, Metric: 97% on 700 instructions)
- [Octo](https://vlabench.dev/entity/octo/): Open transformer policy trained on 800k Open X-Embodiment trajectories and tested across 9 robot platforms. (Status: arXiv 2024, Metric: 800k trajectories)
- [π0](https://vlabench.dev/entity/pi0/): Flow-matching VLA from Physical Intelligence, pretrained on 7 robot configurations and 68 tasks at up to 50 Hz. (Status: arXiv preprint, Metric: 7 robots; 68 tasks)
- [SmolVLA](https://vlabench.dev/entity/smolvla/): Hugging Face VLA with 450M parameters and a flow-matching action expert, released with code, weights, and data. (Status: arXiv 2025, Metric: 450M parameters)
- [GR00T N1](https://vlabench.dev/entity/groot-n1/): NVIDIA humanoid VLA. Public GR00T-N1-2B has 2.2B parameters, and the action module runs at 120 Hz. (Status: Open foundation model, Metric: 2.2B parameters)
- [RDT-1B](https://vlabench.dev/entity/rdt-1b/): Tsinghua diffusion transformer with 1.2B parameters, pretrained on 46 datasets and more than 1M episodes. (Status: Project release, Metric: 1.2B; 1M+ episodes)
- [Open X-Embodiment](https://vlabench.dev/entity/open-x-embodiment/): Cross-lab robot dataset with 1M+ trajectories, 22 embodiments, and 60 source datasets stored as RLDS. (Status: Public dataset, Metric: 1M+ traj.; 22 robots)
- [LIBERO](https://vlabench.dev/entity/libero/): Lifelong manipulation benchmark with 130 language-conditioned tasks in Spatial, Object, Goal, and 100-task suites. (Status: Public benchmark, Metric: 130 tasks; 4 suites)
- [CALVIN](https://vlabench.dev/entity/calvin/): Freiburg long-horizon benchmark: 4 indoor environments, about 24 hours of play data, and 20K language directives. (Status: Public benchmark, Metric: 4 envs; 20K directives)
- [SimplerEnv](https://vlabench.dev/entity/simplerenv/): Real-to-sim policy eval with over 1500 paired sim-and-real trials across two embodiments and eight task families. (Status: PMLR v270, Metric: 1500+ paired evals)
- [DROID](https://vlabench.dev/entity/droid/): In-the-wild Franka dataset: 76k trajectories, 350 hours, 564 scenes, and 86 tasks across 13 institutions. (Status: Public dataset, Metric: 76k traj.; 350 hours)