Compare AI models

Specifications
Parameters
8B/70B16B
Context window
8k8k
Benchmarks
HumanEval
81.7%48.1%
MATH
50.4%45.3%
MMLU
82%70%
GSM8K
93%77.4%
Overview
CompanyMetaMoonshot AI
Release dateApr 18 2024Feb 22 2025
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

LLaMA 3 (8B/70B) leads Moonlight on 4 of the 4 benchmarks they both report (HumanEval, MATH, MMLU, GSM8K). LLaMA 3 (8B/70B) shipped 310 days before Moonlight, so benchmark comparisons should account for the intervening progress.

LLaMA 3 (8B/70B) has 8B/70B parameters, while Moonlight has 16B. Context windows are 8k (LLaMA 3 (8B/70B)) vs 8k (Moonlight).

On HumanEval, LLaMA 3 (8B/70B) leads at 81.7% vs Moonlight at 48.1%. On MATH, LLaMA 3 (8B/70B) leads at 50.4% vs Moonlight at 45.3%. On MMLU, LLaMA 3 (8B/70B) leads at 82% vs Moonlight at 70%. On GSM8K, LLaMA 3 (8B/70B) leads at 93% vs Moonlight at 77.4%.