AlphaFold 2

Training compute

3×10²¹ FLOP

Parameters

93M

Published

Nov 30, 2020

AlphaFold 2 is an AI model developed by DeepMind (United Kingdom), first published in November 2020. It works in the biology domain, on tasks such as protein folding prediction and proteins.

Training it took an estimated 3×10²¹ FLOP of compute (estimation method: hardware). The model has 93,000,000 parameters. It was trained on roughly 5.7B datapoints. Training ran on Google TPU v3 for about 264 hours. The compute alone is estimated at $4K in 2023 dollars.

Access: Open weights (unrestricted). Its weights are openly available. The reference paper has 31,909 citations. Epoch AI rates the confidence of this record as likely.

Full record

Organization: DeepMind
Country of organization: United Kingdom
Domain: Biology
Task: Protein folding prediction, Proteins
Training compute: 3×10²¹ FLOP
Compute estimation method: Hardware
Parameters: 93,000,000
Dataset size: 5.7B
Training hardware: Google TPU v3
Training time: 264 h
Training cost (2023 USD): $4K
Numerical format: BF16
Model accessibility: Open weights (unrestricted)
Open weights: Yes
Citations: 31,909
Epoch confidence: Likely