AI models
OpenVLA
Training compute
1.1×10²³ FLOP
Parameters
7.2B
Published
Jun 13, 2024
OpenVLA is an AI model developed by Stanford University, University of California (UC) Berkeley, Toyota Research Institute, Google DeepMind, Massachusetts Institute of Technology (MIT) and Physical Intelligence (United States), first published in June 2024. It works in the robotics, vision and language domain, on tasks such as robotic manipulation.
Training it took an estimated 1.1×10²³ FLOP of compute (estimation method: hardware). The model has 7,188,100,000 parameters. Training ran on 64 NVIDIA A100 for about 336 hours.
Access: Open weights (unrestricted). Its weights are openly available. It is built on top of Llama 2-7B. The reference paper has 2,123 citations. Epoch AI rates the confidence of this record as confident.
Full record
- Organization
- Stanford University, University of California (UC) Berkeley, Toyota Research Institute, Google DeepMind, Massachusetts Institute of Technology (MIT), Physical Intelligence
- Country of organization
- United States
- Domain
- Robotics, Vision, Language
- Task
- Robotic manipulation
- Training compute
- 1.1×10²³ FLOP
- Compute estimation method
- Hardware
- Parameters
- 7,188,100,000
- Training hardware
- NVIDIA A100
- Chips used
- 64
- Training time
- 336 h
- Training power draw
- 50.5 kW
- Numerical format
- BF16
- Model accessibility
- Open weights (unrestricted)
- Open weights
- Yes
- Base model
- Llama 2-7B
- Citations
- 2,123
- Epoch confidence
- Confident
SourceEpoch AI, 'AI Models'. Published online at epoch.ai. Retrieved 2026-07-29 from https://epoch.ai/data/ai-models. Licensed under CC BY 4.0.
← All ai models