model · model--openvla-7b

OpenVLA 7B

Open 7B VLA trained on the Open X-Embodiment mixture.

Identity and state

PublicationReleased
DeploymentResearch
Canonical claims11
Last reviewed2026-08-29

Reviewed relationships

StateRelated entityRelationScopeBoundary
VerifiedOpen X-Embodimenttrained onpublished training mixtureThe paper states 970k real-robot episodes from Open X-Embodiment.

Canonical evidence claims

Claims are projected from authored tables. A dash remains an explicit missing value; it is never filled from a neighboring product or name match.

MetricValueClass
Code licenseproject and training code published; see repository [foundation-models-m2]policy-model
Kindopen VLA on Llama 2 + SigLIP/DINOv2 [foundation-models-r2]policy-model
PaperarXiv:2406.09246 [foundation-models-r2]policy-model
Parameter count7B [foundation-models-r2]policy-model
Public APIpolicy-model
Public benchmark tablepaper tables only [foundation-models-r2]policy-model
Public datearXiv 2024-06-13; CoRL 2025 [foundation-models-r2]policy-model
Published latency / ratepolicy-model
Stated embodiments in the sourceOpen X-Embodiment mixture (many arms) [foundation-models-r2]policy-model
Stated training data970k real-robot episodes from Open X-Embodiment [foundation-models-r2]policy-model
WeightsHF openvla/openvla-7b; backbone is Llama 2 (Community License) [foundation-models-m2] [foundation-models-r2]policy-model

Source ledger

What is not inferred

No fuzzy entity matching, adjacent-column borrowing, universal compatibility, commercial maturity, or checkpoint-to-product identity. Absence of a relation means unreviewed or unsupported here—not proven incompatibility.