Qwen-14B

Training compute

2.5×10²³ FLOP

Parameters

14B

Published

Sep 28, 2023

Qwen-14B is an AI model developed by Alibaba (China), first published in September 2023. It works in the language domain, on tasks such as language modeling/generation.

Training it took an estimated 2.5×10²³ FLOP of compute (estimation method: operation counting). The model has 14,000,000,000 parameters. It was trained on roughly 3T datapoints.

Access: Open weights (restricted use). Its weights are openly available. The reference paper has 3,775 citations. Epoch AI rates the confidence of this record as confident.

Full record

Organization: Alibaba
Country of organization: China
Domain: Language
Task: Language modeling/generation
Training compute: 2.5×10²³ FLOP
Compute estimation method: Operation counting
Parameters: 14,000,000,000
Dataset size: 3T
Model accessibility: Open weights (restricted use)
Open weights: Yes
Citations: 3,775
Epoch confidence: Confident

Benchmark results

01	Epoch Capabilities Index	ECI Score	112.8
02	BoolQ	Score	0.86
03	ARC AI2	Challenge score	0.84
04	PIQA	Score	0.8
05	LAMBADA	Score	0.71
06	MMLU	EM	0.66
07	GSM8K	EM	0.61
08	BIG-Bench Hard	Average	0.53