Chinchilla

Training compute

5.8×10²³ FLOP

Parameters

70B

Published

Mar 29, 2022

Chinchilla is an AI model developed by DeepMind (United Kingdom), first published in March 2022. It works in the language domain, on tasks such as language modeling.

Training it took an estimated 5.8×10²³ FLOP of compute (estimation method: reported). The model has 70,000,000,000 parameters. It was trained on roughly 1.4T datapoints. Training ran on Google TPU v4,Google TPU v3.

Access: Unreleased. Its weights are not openly released. The reference paper has 3,188 citations. Epoch AI rates the confidence of this record as confident.

Full record

Organization: DeepMind
Country of organization: United Kingdom
Domain: Language
Task: Language modeling
Training compute: 5.8×10²³ FLOP
Compute estimation method: Reported
Parameters: 70,000,000,000
Dataset size: 1.4T
Training hardware: Google TPU v4, Google TPU v3
Numerical format: BF16
Model accessibility: Unreleased
Open weights: No
Citations: 3,188
Epoch confidence: Confident