Compare AI models

Specifications
Parameters
16B14.3B
Context window
8k32k
Benchmarks
HumanEval
48.1%34.2%
MMLU
70%62.5%
GSM8K
77.4%61.5%
Overview
CompanyMoonshot AIQwen
Release dateFeb 22 2025Mar 28 2024
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

Moonlight leads Qwen1.5-MoE on 3 of the 3 benchmarks they both report (HumanEval, MMLU, GSM8K). Qwen1.5-MoE shipped 331 days before Moonlight, so benchmark comparisons should account for the intervening progress.

Moonlight has 16B parameters, while Qwen1.5-MoE has 14.3B. Context windows are 8k (Moonlight) vs 32k (Qwen1.5-MoE).

On HumanEval, Moonlight leads at 48.1% vs Qwen1.5-MoE at 34.2%. On MMLU, Moonlight leads at 70% vs Qwen1.5-MoE at 62.5%. On GSM8K, Moonlight leads at 77.4% vs Qwen1.5-MoE at 61.5%.