Compare AI models

Specifications
Parameters
16B16B
Context window
4k8k
Benchmarks
HumanEval
45.7%48.1%
MATH
15.2%45.3%
MMLU
47.2%70%
GSM8K
62.2%77.4%
Overview
CompanyDeepSeekMoonshot AI
Release dateJan 11 2024Feb 22 2025
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

Moonlight leads DeepSeek-MoE on 4 of the 4 benchmarks they both report (HumanEval, MATH, MMLU, GSM8K). DeepSeek-MoE shipped 408 days before Moonlight, so benchmark comparisons should account for the intervening progress.

DeepSeek-MoE has 16B parameters, while Moonlight has 16B. Context windows are 4k (DeepSeek-MoE) vs 8k (Moonlight).

On HumanEval, Moonlight leads at 48.1% vs DeepSeek-MoE at 45.7%. On MATH, Moonlight leads at 45.3% vs DeepSeek-MoE at 15.2%. On MMLU, Moonlight leads at 70% vs DeepSeek-MoE at 47.2%. On GSM8K, Moonlight leads at 77.4% vs DeepSeek-MoE at 62.2%.