Compare AI models

Specifications
Parameters
236B70B
Context window
128k128k
API pricing
Cheapest input
—$0.10DeepInfra
Cheapest output
—$0.32DeepInfra
—$36.00DeepInfra · turbo · fp8
Benchmarks
HumanEval
84.76%88.4%
MATH
71.02%77%
Overview
CompanyDeepSeekMeta
Release dateJun 28 2024Dec 6 2024
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

LLaMA 3.3 leads DeepSeek V2-0628 on 2 of the 2 benchmarks they both report (HumanEval, MATH). DeepSeek V2-0628 shipped 161 days before LLaMA 3.3, so benchmark comparisons should account for the intervening progress.

DeepSeek V2-0628 has 236B parameters, while LLaMA 3.3 has 70B. Context windows are 128k (DeepSeek V2-0628) vs 128k (LLaMA 3.3).

On HumanEval, LLaMA 3.3 leads at 88.4% vs DeepSeek V2-0628 at 84.76%. On MATH, LLaMA 3.3 leads at 77% vs DeepSeek V2-0628 at 71.02%.