Compare AI models

Specifications
Parameters
—16B
Context window
200k8k
Benchmarks
HumanEval
88.1%48.1%
GPQA Diamond
41.6%35.2%
MMLU
80.9%70%
Overview
CompanyAnthropicMoonshot AI
Release dateOct 22 2024Feb 22 2025
AccessClosedOpen Weight
Model detailsView modelView model

Frequently asked questions

Claude 3.5 Haiku leads Moonlight on 3 of the 3 benchmarks they both report (HumanEval, GPQA Diamond, MMLU). Claude 3.5 Haiku shipped 123 days before Moonlight, so benchmark comparisons should account for the intervening progress.

Context windows are 200k (Claude 3.5 Haiku) vs 8k (Moonlight). Claude 3.5 Haiku is closed, while Moonlight is open weight.

On HumanEval, Claude 3.5 Haiku leads at 88.1% vs Moonlight at 48.1%. On GPQA Diamond, Claude 3.5 Haiku leads at 41.6% vs Moonlight at 35.2%. On MMLU, Claude 3.5 Haiku leads at 80.9% vs Moonlight at 70%.