Compare AI models

Specifications
Parameters
7B/13B/33B/65B14B
Context window
2k2k
Benchmarks
HumanEval
23.7%32.3%
MATH
10.6%24.8%
MMLU
63.4%66.3%
GSM8K
50.9%61.3%
Overview
CompanyMetaQwen
Release dateFeb 24 2023Sep 25 2023
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

Qwen-14B leads LLaMA 1 on 4 of the 4 benchmarks they both report (HumanEval, MATH, MMLU, GSM8K). LLaMA 1 shipped 213 days before Qwen-14B, so benchmark comparisons should account for the intervening progress.

LLaMA 1 has 7B/13B/33B/65B parameters, while Qwen-14B has 14B. Context windows are 2k (LLaMA 1) vs 2k (Qwen-14B).

On HumanEval, Qwen-14B leads at 32.3% vs LLaMA 1 at 23.7%. On MATH, Qwen-14B leads at 24.8% vs LLaMA 1 at 10.6%. On MMLU, Qwen-14B leads at 66.3% vs LLaMA 1 at 63.4%. On GSM8K, Qwen-14B leads at 61.3% vs LLaMA 1 at 50.9%.