Compare AI models

Specifications
Parameters
16B8B/70B/405B
Context window
32k128k
API pricing
Cheapest input
—$0.02DeepInfra
Cheapest output
—$0.04DeepInfra
—$6.00DeepInfra · fp8
Benchmarks
HumanEval
57.3%89%
MATH
27.9%73.8%
MMLU
55.7%87.3%
GSM8K
72%96.8%
Overview
CompanyDeepSeekMeta
Release dateMay 16 2024Jul 23 2024
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

LLaMA 3.1 leads DeepSeek V2 Lite on 4 of the 4 benchmarks they both report (HumanEval, MATH, MMLU, GSM8K). DeepSeek V2 Lite shipped 68 days before LLaMA 3.1, so benchmark comparisons should account for the intervening progress.

DeepSeek V2 Lite has 16B parameters, while LLaMA 3.1 has 8B/70B/405B. Context windows are 32k (DeepSeek V2 Lite) vs 128k (LLaMA 3.1).

On HumanEval, LLaMA 3.1 leads at 89% vs DeepSeek V2 Lite at 57.3%. On MATH, LLaMA 3.1 leads at 73.8% vs DeepSeek V2 Lite at 27.9%. On MMLU, LLaMA 3.1 leads at 87.3% vs DeepSeek V2 Lite at 55.7%. On GSM8K, LLaMA 3.1 leads at 96.8% vs DeepSeek V2 Lite at 72%.