Compare AI models

Specifications
Parameters
16B—
Context window
8k128k
Benchmarks
HumanEval
48.1%75.7%
MATH
45.3%57.9%
MMLU
70%81.9%
GSM8K
77.4%90.9%
Overview
CompanyMoonshot AIZ.ai
Release dateFeb 22 2025Jun 5 2024
AccessOpen WeightClosed
Model detailsView modelView model

Frequently asked questions

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

Context windows are 8k (Moonlight) vs 128k (GLM-4-Air). Moonlight is open weight, while GLM-4-Air is closed.

On HumanEval, GLM-4-Air leads at 75.7% vs Moonlight at 48.1%. On MATH, GLM-4-Air leads at 57.9% vs Moonlight at 45.3%. On MMLU, GLM-4-Air leads at 81.9% vs Moonlight at 70%. On GSM8K, GLM-4-Air leads at 90.9% vs Moonlight at 77.4%.