Compare AI models

Specifications
Parameters
16B110B
Context window
8k32k
Benchmarks
HumanEval
48.1%52.4%
MATH
45.3%49.6%
MMLU
70%80.4%
GSM8K
77.4%85.4%
Overview
CompanyMoonshot AIQwen
Release dateFeb 22 2025Apr 25 2024
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

Qwen1.5-110B leads Moonlight on 4 of the 4 benchmarks they both report (HumanEval, MATH, MMLU, GSM8K). Qwen1.5-110B shipped 303 days before Moonlight, so benchmark comparisons should account for the intervening progress.

Moonlight has 16B parameters, while Qwen1.5-110B has 110B. Context windows are 8k (Moonlight) vs 32k (Qwen1.5-110B).

On HumanEval, Qwen1.5-110B leads at 52.4% vs Moonlight at 48.1%. On MATH, Qwen1.5-110B leads at 49.6% vs Moonlight at 45.3%. On MMLU, Qwen1.5-110B leads at 80.4% vs Moonlight at 70%. On GSM8K, Qwen1.5-110B leads at 85.4% vs Moonlight at 77.4%.