Compare AI models

Specifications
Parameters
16B314B
Context window
8k8k
Benchmarks
HumanEval
48.1%63.2%
MATH
45.3%23.9%
MMLU
70%73%
GSM8K
77.4%62.9%
Overview
CompanyMoonshot AISpaceXSI
Release dateFeb 22 2025Nov 3 2023
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

Moonlight and Grok‑1 are evenly matched across the 4 benchmarks they both report (HumanEval, MATH, MMLU, GSM8K). Grok‑1 shipped 477 days before Moonlight, so benchmark comparisons should account for the intervening progress.

Moonlight has 16B parameters, while Grok‑1 has 314B. Context windows are 8k (Moonlight) vs 8k (Grok‑1).

On HumanEval, Grok‑1 leads at 63.2% vs Moonlight at 48.1%. On MATH, Moonlight leads at 45.3% vs Grok‑1 at 23.9%. On MMLU, Grok‑1 leads at 73% vs Moonlight at 70%. On GSM8K, Moonlight leads at 77.4% vs Grok‑1 at 62.9%.