DEPLOYDatabase

Brain

OpenVLA

An academic open-source vision-language-action model at 7B parameters. Multi-developer collaboration across Stanford, UC Berkeley, Toyota Research Institute, and Google DeepMind. Listed here as a research-model entity; architectural details should be firmed against the OpenVLA paper at build.

Research model · Maturity: Research · Open source

Machine-readable surfaces

Architecture

7B-parameter VLA. Architecture and training-data details pending firmer primary-source confirmation (OpenVLA paper at openvla.github.io and arXiv).

Key facts

Thin-record depth note
Architecture, parameter count, training-data composition should be firmed against the OpenVLA paper at openvla.github.io / arXiv. Entity carries verified-vs-claimed framing until firmed.
Multi-developer collaboration
Stanford + UC Berkeley + Toyota Research Institute + Google DeepMind (only Google DeepMind is attached as a registry developer; the academic institutions are not registry Companies)
Powered platforms
Research robot platforms; not a commercial-deployment brain
7B VLA OpenX pretrain (primary OpenVLA site)
The OpenVLA project page describes a 7B-parameter open-source vision-language-action model pretrained on 970k robot episodes from the Open X-Embodiment dataset. Architecture listed: fused SigLIP + DinoV2 visual encoder, projector into a Llama 2 7B backbone that predicts tokenized actions decoded to continuous robot actions. Training cited on 64 A100 GPUs for 15 days. Checkpoints and PyTorch pipeline are open-sourced on HuggingFace (OpenVLA)
Authors and eval framing (primary OpenVLA / arXiv)
Authors listed include Moo Jin Kim, Karl Pertsch, Siddharth Karamcheti (equal contribution) with affiliations spanning Stanford, UC Berkeley, Toyota Research Institute, Google DeepMind, Physical Intelligence, and MIT. The page says OpenVLA sets a new state of the art for generalist robot manipulation policies versus RT-1-X and Octo in their evaluations, and can be adapted via parameter-efficient fine-tuning (LoRA matching full fine-tune while updating about 1.4% of parameters). Treat SOTA wording as authors' reported eval claim (OpenVLA; arXiv:2406.09246)
Open model distribution
The OpenVLA project publishes model checkpoints, code, and training materials for an open-source 7B vision-language-action model for robot manipulation (OpenVLA).
Inference and fine-tuning resources
The official OpenVLA repository documents loading the model for inference and adapting it to new robot tasks through fine-tuning (OpenVLA repository).

Developed by (1)

Mentioned on (3)

Registry pages whose text names OpenVLA.

Companies (1)

Walden Robotics

AI systems (2)

GraspVLA (Galbot), RDT-1B

Common questions

What is OpenVLA?
An academic open-source vision-language-action model at 7B parameters. Multi-developer collaboration across Stanford, UC Berkeley, Toyota Research Institute, and Google DeepMind. Listed here as a research-model entity; architectural details should be firmed against the OpenVLA paper at build.
Who developed OpenVLA?
OpenVLA is credited to Google DeepMind on the DEPLOY registry. Each developer attribution is verified via primary sources.
Is OpenVLA open source?
Yes. OpenVLA is recorded as open-source / open-weights on the DEPLOY registry, meaning model weights or source code are publicly available.
What type of AI is OpenVLA?
OpenVLA is a research model, built on a 7B-parameter VLA. Architecture and training-data details pending firmer primary-source confirmation (OpenVLA paper at openvla.github.io and arXiv). architecture on the DEPLOY registry.
What is OpenVLA's maturity stage?
OpenVLA is at the research stage on the DEPLOY maturity ladder. Research stage means active development without commercial deployments on file.
Which robots run on OpenVLA?
No robot models on the DEPLOY registry are recorded as running OpenVLA. DEPLOY wires brain-to-model connections only when the wiring is verifiable from primary sources; absence may reflect pre-deployment or unverified manufacturer claims.

Sources (3)

  1. OpenVLA project page (firm at build) · https://openvla.github.io/
  2. https://arxiv.org/abs/2406.09246
  3. https://github.com/openvla/openvla
Methodology: Verified · 3 sources (no primary) · last reviewed 2026-10-09

Verification posture

Verified

Low confidence

Review state

Stable

Last reviewed 2026-10-09

Maturity + lifecycle

Maturity stage: research

Sources by quality tier

1
unclassified
Unclassified source
1
preprint
Preprint
1
code-repository
Code repository

The framework is documented at /methodology. Corrections at /corrections. Reviewer: DEPLOY editorial team.

Methodology surface for OpenVLA.

Canonical ID 8ca71205-71ca-4036-8f26-1bb35217fa35