Compare AI models

Specifications
Parameters
16B6B
Context window
8k2k
Benchmarks
HumanEval
48.1%0%
MATH
45.3%3.1%
MMLU
70%25.2%
GSM8K
77.4%1.5%
Overview
CompanyMoonshot AIZ.ai
Release dateFeb 22 2025Mar 14 2023
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

Moonlight leads ChatGLM-6B on 4 of the 4 benchmarks they both report (HumanEval, MATH, MMLU, GSM8K). ChatGLM-6B shipped 711 days before Moonlight, so benchmark comparisons should account for the intervening progress.

Moonlight has 16B parameters, while ChatGLM-6B has 6B. Context windows are 8k (Moonlight) vs 2k (ChatGLM-6B).

On HumanEval, Moonlight leads at 48.1% vs ChatGLM-6B at 0%. On MATH, Moonlight leads at 45.3% vs ChatGLM-6B at 3.1%. On MMLU, Moonlight leads at 70% vs ChatGLM-6B at 25.2%. On GSM8K, Moonlight leads at 77.4% vs ChatGLM-6B at 1.5%.