RWKV-4 14B

Training compute

2.8×10²² FLOP

Parameters

14B

Published

May 22, 2023

RWKV-4 14B is an AI model developed by RWKV Foundation, first published in May 2023. It works in the language domain, on tasks such as language modeling.

Training it took an estimated 2.8×10²² FLOP of compute (estimation method: operation counting). The model has 14,000,000,000 parameters. It was trained on roughly 330B datapoints. Training ran on NVIDIA A100 SXM4 80 GB.

The reference paper has 1,042 citations. Epoch AI rates the confidence of this record as confident.

Full record

Organization: RWKV Foundation
Domain: Language
Task: Language modeling
Training compute: 2.8×10²² FLOP
Compute estimation method: Operation counting
Parameters: 14,000,000,000
Dataset size: 330B
Training hardware: NVIDIA A100 SXM4 80 GB
Citations: 1,042
Epoch confidence: Confident