Compare AI models

Specifications
Parameters
16B—
Context window
8k200k
Benchmarks
HumanEval
48.1%97.6%
GPQA Diamond
35.2%77.2%
SimpleQA
5.6%13.8%
MMLU
70%86.9%
Overview
CompanyMoonshot AIOpenAI
Release dateFeb 22 2025Jan 31 2025
AccessOpen WeightClosed
Model detailsView modelView model

Frequently asked questions

o3-mini-high leads Moonlight on 4 of the 4 benchmarks they both report (HumanEval, GPQA Diamond, SimpleQA, MMLU). o3-mini-high shipped 22 days before Moonlight, so benchmark comparisons should account for the intervening progress.

Context windows are 8k (Moonlight) vs 200k (o3-mini-high). Moonlight is open weight, while o3-mini-high is closed.

On HumanEval, o3-mini-high leads at 97.6% vs Moonlight at 48.1%. On GPQA Diamond, o3-mini-high leads at 77.2% vs Moonlight at 35.2%. On SimpleQA, o3-mini-high leads at 13.8% vs Moonlight at 5.6%. On MMLU, o3-mini-high leads at 86.9% vs Moonlight at 70%.