Compare AI models

Specifications
Parameters
236B16B
Context window
128k8k
Benchmarks
LiveCodeBench
32.5%14.3%
HumanEval
81.1%48.1%
MATH
53.9%45.3%
MMLU
77.8%70%
GSM8K
92.2%77.4%
Overview
CompanyDeepSeekMoonshot AI
Release dateMay 6 2024Feb 22 2025
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

DeepSeek V2 leads Moonlight on 5 of the 5 benchmarks they both report. DeepSeek V2 shipped 292 days before Moonlight, so benchmark comparisons should account for the intervening progress.

DeepSeek V2 has 236B parameters, while Moonlight has 16B. Context windows are 128k (DeepSeek V2) vs 8k (Moonlight).

On LiveCodeBench, DeepSeek V2 leads at 32.5% vs Moonlight at 14.3%. On HumanEval, DeepSeek V2 leads at 81.1% vs Moonlight at 48.1%. On MATH, DeepSeek V2 leads at 53.9% vs Moonlight at 45.3%. On MMLU, DeepSeek V2 leads at 77.8% vs Moonlight at 70%. On GSM8K, DeepSeek V2 leads at 92.2% vs Moonlight at 77.4%.