AI models
Tranception
Training compute
7.2×10²¹ FLOP
Parameters
700M
Published
May 27, 2022
Tranception is an AI model developed by University of Oxford, Harvard Medical School and Cohere (United Kingdom, United States and Canada), first published in May 2022. It works in the biology domain, on tasks such as proteins and protein pathogenicity prediction.
Training it took an estimated 7.2×10²¹ FLOP of compute (estimation method: hardware). The model has 700,000,000 parameters. It was trained on roughly 48.2B datapoints. Training ran on 64 NVIDIA A100 for about 336 hours. The compute alone is estimated at $15K in 2023 dollars.
Access: Open weights (unrestricted). Its weights are openly available. The reference paper has 243 citations. Epoch AI rates the confidence of this record as confident.
Full record
- Organization
- University of Oxford, Harvard Medical School, Cohere
- Country of organization
- United Kingdom, United States, Canada
- Domain
- Biology
- Task
- Proteins, Protein pathogenicity prediction
- Training compute
- 7.2×10²¹ FLOP
- Compute estimation method
- Hardware
- Parameters
- 700,000,000
- Dataset size
- 48.2B
- Training hardware
- NVIDIA A100
- Chips used
- 64
- Training time
- 336 h
- Chip-hours
- 21.5K
- Training power draw
- 51.4 kW
- Training cost (2023 USD)
- $15K
- Model accessibility
- Open weights (unrestricted)
- Open weights
- Yes
- Citations
- 243
- 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