Compare AI models

Specifications
Parameters
7B16B
Context window
4k8k
Benchmarks
MATH
51.7%45.3%
GSM8K
88.2%77.4%
Overview
CompanyDeepSeekMoonshot AI
Release dateFeb 5 2024Feb 22 2025
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

DeepSeek-Math leads Moonlight on 2 of the 2 benchmarks they both report (MATH, GSM8K). DeepSeek-Math shipped 383 days before Moonlight, so benchmark comparisons should account for the intervening progress.

DeepSeek-Math has 7B parameters, while Moonlight has 16B. Context windows are 4k (DeepSeek-Math) vs 8k (Moonlight).

On MATH, DeepSeek-Math leads at 51.7% vs Moonlight at 45.3%. On GSM8K, DeepSeek-Math leads at 88.2% vs Moonlight at 77.4%.