Compare AI models

Specifications
Parameters
16B—
Context window
8k128k
Benchmarks
HumanEval
48.1%83.7%
MATH
45.3%64.3%
GPQA Diamond
35.2%42.5%
MMLU
70%84.7%
Overview
CompanyMoonshot AIOpenAI
Release dateFeb 22 2025Nov 6 2023
AccessOpen WeightClosed
Model detailsView modelView model

Frequently asked questions

GPT-4 Turbo leads Moonlight on 4 of the 4 benchmarks they both report (HumanEval, MATH, GPQA Diamond, MMLU). GPT-4 Turbo shipped 474 days before Moonlight, so benchmark comparisons should account for the intervening progress.

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

On HumanEval, GPT-4 Turbo leads at 83.7% vs Moonlight at 48.1%. On MATH, GPT-4 Turbo leads at 64.3% vs Moonlight at 45.3%. On GPQA Diamond, GPT-4 Turbo leads at 42.5% vs Moonlight at 35.2%. On MMLU, GPT-4 Turbo leads at 84.7% vs Moonlight at 70%.