DISTRO

Training compute

7.1×10²⁰ FLOP

Parameters

1.2B

Published

Aug 26, 2024

DISTRO is an AI model developed by Nous Research (United States), first published in August 2024. It works in the language domain, on tasks such as language modeling/generation, chat and question answering.

Training it took an estimated 7.1×10²⁰ FLOP of compute (estimation method: operation counting,hardware). The model has 1,200,000,000 parameters. It was trained on roughly 100B datapoints. Training ran on 32 NVIDIA H100 SXM5 80GB for about 19.8 hours.

Epoch AI rates the confidence of this record as confident.

Full record

Organization: Nous Research
Country of organization: United States
Domain: Language
Task: Language modeling/generation, Chat, Question answering
Training compute: 7.1×10²⁰ FLOP
Compute estimation method: Operation counting, Hardware
Parameters: 1,200,000,000
Dataset size: 100B
Training hardware: NVIDIA H100 SXM5 80GB
Chips used: 32
Training time: 20 h
Training power draw: 44.2 kW
Epoch confidence: Confident