GPT3-6.7B (rerun of original)

Training compute

1.2×10²² FLOP

Parameters

6.7B

Published

May 28, 2020

GPT3-6.7B (rerun of original) is an AI model developed by Microsoft and OpenAI (United States), first published in May 2020. It works in the language domain, on tasks such as language modeling/generation.

Training it took an estimated 1.2×10²² FLOP of compute (estimation method: operation counting). The model has 6,700,000,000 parameters. It was trained on roughly 300B datapoints. Training ran on NVIDIA V100.

Access: Unreleased. Its weights are not openly released. The reference paper has 262 citations. Epoch AI rates the confidence of this record as confident.

Full record

Organization: Microsoft, OpenAI
Country of organization: United States
Domain: Language
Task: Language modeling/generation
Training compute: 1.2×10²² FLOP
Compute estimation method: Operation counting
Parameters: 6,700,000,000
Dataset size: 300B
Training hardware: NVIDIA V100
Numerical format: FP32
Model accessibility: Unreleased
Open weights: No
Citations: 262
Epoch confidence: Confident