Compare AI models

Specifications
Parameters
16B9B
Context window
8k128k
Benchmarks
HumanEval
48.1%71.8%
MATH
45.3%50.6%
MMLU
70%72.4%
GSM8K
77.4%79.6%
Overview
CompanyMoonshot AIZ.ai
Release dateFeb 22 2025Jun 5 2024
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

GLM-4-9B leads Moonlight on 4 of the 4 benchmarks they both report (HumanEval, MATH, MMLU, GSM8K). GLM-4-9B shipped 262 days before Moonlight, so benchmark comparisons should account for the intervening progress.

Moonlight has 16B parameters, while GLM-4-9B has 9B. Context windows are 8k (Moonlight) vs 128k (GLM-4-9B).

On HumanEval, GLM-4-9B leads at 71.8% vs Moonlight at 48.1%. On MATH, GLM-4-9B leads at 50.6% vs Moonlight at 45.3%. On MMLU, GLM-4-9B leads at 72.4% vs Moonlight at 70%. On GSM8K, GLM-4-9B leads at 79.6% vs Moonlight at 77.4%.