Compare AI models

Specifications
Parameters
16B7B/13B/70B
Context window
32k4k
Benchmarks
HumanEval
57.3%29.9%
MATH
27.9%13.5%
MMLU
55.7%68.9%
GSM8K
72%56.8%
Overview
CompanyDeepSeekMeta
Release dateMay 16 2024Jul 18 2023
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

DeepSeek V2 Lite leads LLaMA 2 on 3 of the 4 benchmarks they both report (HumanEval, MATH, MMLU, GSM8K). LLaMA 2 shipped 303 days before DeepSeek V2 Lite, so benchmark comparisons should account for the intervening progress.

DeepSeek V2 Lite has 16B parameters, while LLaMA 2 has 7B/13B/70B. Context windows are 32k (DeepSeek V2 Lite) vs 4k (LLaMA 2).

On HumanEval, DeepSeek V2 Lite leads at 57.3% vs LLaMA 2 at 29.9%. On MATH, DeepSeek V2 Lite leads at 27.9% vs LLaMA 2 at 13.5%. On MMLU, LLaMA 2 leads at 68.9% vs DeepSeek V2 Lite at 55.7%. On GSM8K, DeepSeek V2 Lite leads at 72% vs LLaMA 2 at 56.8%.