Minerva (540B)
Minerva (540B) is an AI model developed by Google (United States), first published in June 2022. It works in the language domain, on tasks such as quantitative reasoning, mathematical reasoning, language modeling/generation and question answering. It counts among the frontier models: the systems trained with the most compute of their moment.
Training it took an estimated 2.7×10²⁴ FLOP of compute (estimation method: hardware). The model has 540,350,000,000 parameters. It was trained on roughly 26B datapoints. Training ran on 1,024 Google TPU v4 for about 696 hours.
Access: Unreleased. Its weights are not openly released. It is built on top of PaLM (540B). The reference paper has 1,673 citations. Epoch AI rates the confidence of this record as confident.
- Organization
- Country of organization
- United States
- Domain
- Language
- Task
- Quantitative reasoning, Mathematical reasoning, Language modeling/generation, Question answering
- Training compute
- 2.7×10²⁴ FLOP
- Compute estimation method
- Hardware
- Parameters
- 540,350,000,000
- Dataset size
- 26B
- Training hardware
- Google TPU v4
- Chips used
- 1,024
- Training time
- 696 h
- Chip-hours
- 712.7K
- Training power draw
- 698.4 kW
- Model accessibility
- Unreleased
- Open weights
- No
- Base model
- PaLM (540B)
- Citations
- 1,673
- Epoch confidence
- Confident