Skip to content
RobothecaBeta
Log in

OpenVLA

  • AI model
  • Humanoids and robot learning

OpenVLA is an open-source vision-language-action model for generalist robotic manipulation, built on the Prismatic VLM codebase, with pretrained checkpoints and training and fine-tuning scripts.

  • Custom weights terms
  • Stale

Made by OpenVLA Collaboration.

Source on GitHub: github.com/openvla/openvla (opens in a new tab). Facts read at commit c8f03f4 (opens in a new tab).

Licenses

Code
MIT (opens in a new tab)

Read by Robotheca from LICENSE (opens in a new tab) (MIT) at commit c8f03f4 (opens in a new tab), on 16 Oct 2026.

Weights
Llama 2 Community LicenseTerms of their own, not a standard license

Read on github.com/openvla/openvla/blob/main/README.md (opens in a new tab).

Commercial use
Yes

Read as the Llama 2 Community License (the README; the model card says MIT): commercial use allowed, a separate license above 700 million monthly users, and an acceptable use policy.

Labels

  • Custom weights termsThe weights come under their maker's own terms (Llama 2 Community License), not a standard license: read their conditions before any use.
  • StaleNo commit in 12 months: the newest is from 24 Mar 2025.

What it does

  • Loads the openvla-7b checkpoint through HuggingFace transformers to predict 7-DoF robot actions from a camera image and an instruction.
  • Supports training on arbitrary RLDS-format datasets, including mixtures from the Open X-Embodiment dataset.
  • Scales training with PyTorch FSDP and Flash-Attention, from 1B to 34B parameter models.
  • Provides LoRA fine-tuning via HuggingFace PEFT, as well as full fine-tuning on a single node of 8 GPUs.
  • Includes a script to serve models over a REST API so robot stacks need no powerful on-device compute.
  • Has evaluation instructions for BridgeData V2 WidowX robots and the LIBERO simulation benchmark.

Written by Robotheca from the project's README (opens in a new tab) at commit c8f03f4.

Getting started

Difficulty
Advanced

Maintenance

Status
Stale: no commit in 12 months
Newest commit
24 Mar 2025 on main
Latest release
None published on GitHub

Checked on GitHub on 16 Oct 2026.

Links

Robots on Robotheca that use it

No robot on Robotheca is linked to it yet. Links come from a robot's robot.yaml, from its dependency files read as text, or from Robotheca's editors.

Robotheca lists AI models as a reference: nothing is hosted or run here, and every fact links to where it was read.