T5-11B
T5-11B is an AI model developed by Google (United States), first published in October 2019. It works in the language domain, on tasks such as text autocompletion and language modeling/generation. It counts among the frontier models: the systems trained with the most compute of their moment.
Training it took an estimated 3.3×10²² FLOP of compute (estimation method: reported,operation counting,third-party estimation). The model has 11,000,000,000 parameters. It was trained on roughly 34B datapoints. Training ran on 512 Google TPU v3 for about 482 hours. The compute alone is estimated at $78K in 2023 dollars.
Access: Open weights (unrestricted). Its weights are openly available. The reference paper has 25,683 citations. Epoch AI rates the confidence of this record as confident.
- Organization
- Country of organization
- United States
- Domain
- Language
- Task
- Text autocompletion, Language modeling/generation
- Training compute
- 3.3×10²² FLOP
- Compute estimation method
- Reported, Operation counting, Third-party estimation
- Parameters
- 11,000,000,000
- Dataset size
- 34B
- Training hardware
- Google TPU v3
- Chips used
- 512
- Training time
- 482 h
- Chip-hours
- 246.7K
- Training power draw
- 472.4 kW
- Training cost (2023 USD)
- $78K
- Model accessibility
- Open weights (unrestricted)
- Open weights
- Yes
- Citations
- 25,683
- Epoch confidence
- Confident
| 01 | BoolQ | Score | 0.91 |
| 02 | SuperGLUE | Score | 0.89 |
| 03 | CommonsenseQA 2 | Score | 0.68 |