AI models
AudioGen
Training compute
9.5×10²¹ FLOP
Parameters
1B
Published
Mar 5, 2023
AudioGen is an AI model developed by Meta AI and Hebrew University of Jerusalem (United States and Israel), first published in March 2023. It works in the audio domain, on tasks such as audio generation.
Training it took an estimated 9.5×10²¹ FLOP of compute (estimation method: hardware). The model has 1,000,000,000 parameters. It was trained on roughly 230.4B datapoints. Training ran on NVIDIA A100 for about 168 hours. The compute alone is estimated at $9K in 2023 dollars.
Access: Open weights (non-commercial). Its weights are openly available. The reference paper has 436 citations. Epoch AI rates the confidence of this record as likely.
Full record
- Organization
- Meta AI, Hebrew University of Jerusalem
- Country of organization
- United States, Israel
- Domain
- Audio
- Task
- Audio generation
- Training compute
- 9.5×10²¹ FLOP
- Compute estimation method
- Hardware
- Parameters
- 1,000,000,000
- Dataset size
- 230.4B
- Training hardware
- NVIDIA A100
- Training time
- 168 h
- Training cost (2023 USD)
- $9K
- Model accessibility
- Open weights (non-commercial)
- Open weights
- Yes
- Citations
- 436
- Epoch confidence
- Likely
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