Luminous-base

Training compute

3.2×10²² FLOP

Parameters

13B

Published

Aug 15, 2022

Luminous-base is an AI model developed by Aleph Alpha (Germany), first published in August 2022. It works in the language domain, on tasks such as language modeling/generation.

Training it took an estimated 3.2×10²² FLOP of compute (estimation method: operation counting,hardware). The model has 13,000,000,000 parameters. It was trained on roughly 402B datapoints. Training ran on 128 NVIDIA A100 SXM4 40 GB for about 1.3K hours.

Access: API access. Its weights are not openly released. Epoch AI rates the confidence of this record as confident.

Full record

Organization: Aleph Alpha
Country of organization: Germany
Domain: Language
Task: Language modeling/generation
Training compute: 3.2×10²² FLOP
Compute estimation method: Operation counting, Hardware
Parameters: 13,000,000,000
Dataset size: 402B
Training hardware: NVIDIA A100 SXM4 40 GB
Chips used: 128
Training time: 1,344 h
Chip-hours: 95K
Training power draw: 102.6 kW
Model accessibility: API access
Open weights: No
Epoch confidence: Confident