Compare AI models

Specifications
Parameters
—16B
Context window
100k8k
Benchmarks
HumanEval
71.2%48.1%
MMLU
78.5%70%
GSM8K
88%77.4%
Overview
CompanyAnthropicMoonshot AI
Release dateJul 11 2023Feb 22 2025
AccessClosedOpen Weight
Model detailsView modelView model

Frequently asked questions

Claude 2 leads Moonlight on 3 of the 3 benchmarks they both report (HumanEval, MMLU, GSM8K). Claude 2 shipped 592 days before Moonlight, so benchmark comparisons should account for the intervening progress.

Context windows are 100k (Claude 2) vs 8k (Moonlight). Claude 2 is closed, while Moonlight is open weight.

On HumanEval, Claude 2 leads at 71.2% vs Moonlight at 48.1%. On MMLU, Claude 2 leads at 78.5% vs Moonlight at 70%. On GSM8K, Claude 2 leads at 88% vs Moonlight at 77.4%.