Compare AI models

Specifications
Parameters
9B/27B16B
Context window
8k8k
API pricing
Cheapest input
$0.65NextBit—
Cheapest output
$0.65NextBit—
$163NextBit · int4—
Benchmarks
HumanEval
51.8%48.1%
MATH
42.3%45.3%
MMLU
75.2%70%
GSM8K
74%77.4%
Overview
CompanyGoogleMoonshot AI
Release dateJun 27 2024Feb 22 2025
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

Gemma 2 and Moonlight are evenly matched across the 4 benchmarks they both report (HumanEval, MATH, MMLU, GSM8K). Gemma 2 shipped 240 days before Moonlight, so benchmark comparisons should account for the intervening progress.

Gemma 2 has 9B/27B parameters, while Moonlight has 16B. Context windows are 8k (Gemma 2) vs 8k (Moonlight).

On HumanEval, Gemma 2 leads at 51.8% vs Moonlight at 48.1%. On MATH, Moonlight leads at 45.3% vs Gemma 2 at 42.3%. On MMLU, Gemma 2 leads at 75.2% vs Moonlight at 70%. On GSM8K, Moonlight leads at 77.4% vs Gemma 2 at 74%.