Compare AI models

Specifications
Parameters
7B/13B/33B/65B7B
Context window
2k32k
Benchmarks
HumanEval
23.7%38.4%
MATH
10.6%13.2%
Overview
CompanyMetaMistral
Release dateFeb 24 2023May 22 2024
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

Mistral 7B v0.3 leads LLaMA 1 on 2 of the 2 benchmarks they both report (HumanEval, MATH). LLaMA 1 shipped 453 days before Mistral 7B v0.3, so benchmark comparisons should account for the intervening progress.

LLaMA 1 has 7B/13B/33B/65B parameters, while Mistral 7B v0.3 has 7B. Context windows are 2k (LLaMA 1) vs 32k (Mistral 7B v0.3).

On HumanEval, Mistral 7B v0.3 leads at 38.4% vs LLaMA 1 at 23.7%. On MATH, Mistral 7B v0.3 leads at 13.2% vs LLaMA 1 at 10.6%.