AlphaGo Zero

Training compute

6.5×10²⁰ FLOP

Parameters

46.4M

Published

Oct 18, 2017

AlphaGo Zero is an AI model developed by DeepMind (United Kingdom), first published in October 2017. It works in the games domain, on tasks such as go. It counts among the frontier models: the systems trained with the most compute of their moment.

Training it took an estimated 6.5×10²⁰ FLOP of compute (estimation method: third-party estimation,hardware). The model has 46,400,244 parameters. It was trained on roughly 6.3B datapoints. Training ran on Google TPU v1 for about 480 hours. The compute alone is estimated at $1K in 2023 dollars.

Access: Unreleased. Its weights are not openly released. The reference paper has 10,021 citations. Epoch AI rates the confidence of this record as confident.

Full record

Organization: DeepMind
Country of organization: United Kingdom
Domain: Games
Task: Go
Training compute: 6.5×10²⁰ FLOP
Compute estimation method: Third-party estimation, Hardware
Parameters: 46,400,244
Dataset size: 6.3B
Training hardware: Google TPU v1
Training time: 480 h
Training cost (2023 USD): $1K
Model accessibility: Unreleased
Open weights: No
Citations: 10,021
Epoch confidence: Confident