Compare AI models

Specifications
Parameters
7B/13B/33B/65B6B
Context window
2k32k
Benchmarks
HumanEval
23.7%9.8%
MATH
10.6%6.9%
HellaSwag
84.2%57%
MMLU
63.4%45.2%
GSM8K
50.9%25.9%
Overview
CompanyMetaZ.ai
Release dateFeb 24 2023Jun 25 2023
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

LLaMA 1 leads ChatGLM2-6B on 5 of the 5 benchmarks they both report. LLaMA 1 shipped 121 days before ChatGLM2-6B, so benchmark comparisons should account for the intervening progress.

LLaMA 1 has 7B/13B/33B/65B parameters, while ChatGLM2-6B has 6B. Context windows are 2k (LLaMA 1) vs 32k (ChatGLM2-6B).

On HumanEval, LLaMA 1 leads at 23.7% vs ChatGLM2-6B at 9.8%. On MATH, LLaMA 1 leads at 10.6% vs ChatGLM2-6B at 6.9%. On HellaSwag, LLaMA 1 leads at 84.2% vs ChatGLM2-6B at 57%. On MMLU, LLaMA 1 leads at 63.4% vs ChatGLM2-6B at 45.2%. On GSM8K, LLaMA 1 leads at 50.9% vs ChatGLM2-6B at 25.9%.