Compare AI models

Specifications
Parameters
2B/7B16B
Context window
8k8k
Benchmarks
HumanEval
32.3%48.1%
MATH
24.3%45.3%
MMLU
64.3%70%
GSM8K
46.4%77.4%
Overview
CompanyGoogleMoonshot AI
Release dateFeb 21 2024Feb 22 2025
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

Moonlight leads Gemma 1 on 4 of the 4 benchmarks they both report (HumanEval, MATH, MMLU, GSM8K). Gemma 1 shipped 367 days before Moonlight, so benchmark comparisons should account for the intervening progress.

Gemma 1 has 2B/7B parameters, while Moonlight has 16B. Context windows are 8k (Gemma 1) vs 8k (Moonlight).

On HumanEval, Moonlight leads at 48.1% vs Gemma 1 at 32.3%. On MATH, Moonlight leads at 45.3% vs Gemma 1 at 24.3%. On MMLU, Moonlight leads at 70% vs Gemma 1 at 64.3%. On GSM8K, Moonlight leads at 77.4% vs Gemma 1 at 46.4%.