Baichuan1-7B

Training compute

5×10²² FLOP

Parameters

7B

Published

Jun 1, 2023

Baichuan1-7B is an AI model developed by Baichuan (China), first published in June 2023. It works in the language domain, on tasks such as language modeling/generation and question answering.

Training it took an estimated 5×10²² FLOP of compute (estimation method: operation counting). The model has 7,000,559,616 parameters. It was trained on roughly 1.2T datapoints. Training ran on 1,024 NVIDIA A800 PCIe 40 GB.

Access: Open weights (non-commercial). Its weights are openly available. Epoch AI rates the confidence of this record as confident.

Full record

Organization: Baichuan
Country of organization: China
Domain: Language
Task: Language modeling/generation, Question answering
Training compute: 5×10²² FLOP
Compute estimation method: Operation counting
Parameters: 7,000,559,616
Dataset size: 1.2T
Training hardware: NVIDIA A800 PCIe 40 GB
Chips used: 1,024
Training power draw: 509.7 kW
Model accessibility: Open weights (non-commercial)
Open weights: Yes
Epoch confidence: Confident

Benchmark results

01

Epoch Capabilities Index

ECI Score

90