Compare AI models

Specifications
Parameters
—16B
Context window
32k8k
Benchmarks
HumanEval
67.7%48.1%
MATH
32.6%45.3%
MMLU
71.8%70%
GSM8K
77.9%77.4%
Overview
CompanyGoogleMoonshot AI
Release dateDec 13 2023Feb 22 2025
AccessClosedOpen Weight
Model detailsView modelView model

Frequently asked questions

Gemini 1.0 Pro leads Moonlight on 3 of the 4 benchmarks they both report (HumanEval, MATH, MMLU, GSM8K). Gemini 1.0 Pro shipped 437 days before Moonlight, so benchmark comparisons should account for the intervening progress.

Context windows are 32k (Gemini 1.0 Pro) vs 8k (Moonlight). Gemini 1.0 Pro is closed, while Moonlight is open weight.

On HumanEval, Gemini 1.0 Pro leads at 67.7% vs Moonlight at 48.1%. On MATH, Moonlight leads at 45.3% vs Gemini 1.0 Pro at 32.6%. On MMLU, Gemini 1.0 Pro leads at 71.8% vs Moonlight at 70%. On GSM8K, Gemini 1.0 Pro leads at 77.9% vs Moonlight at 77.4%.