Compare AI models

Specifications
Parameters
—16B
Context window
100k32k
Benchmarks
HumanEval
71.2%57.3%
MMLU
78.5%55.7%
GSM8K
88%72%
Overview
CompanyAnthropicDeepSeek
Release dateJul 11 2023May 16 2024
AccessClosedOpen Weight
Model detailsView modelView model

Frequently asked questions

Claude 2 leads DeepSeek V2 Lite on 3 of the 3 benchmarks they both report (HumanEval, MMLU, GSM8K). Claude 2 shipped 310 days before DeepSeek V2 Lite, so benchmark comparisons should account for the intervening progress.

Context windows are 100k (Claude 2) vs 32k (DeepSeek V2 Lite). Claude 2 is closed, while DeepSeek V2 Lite is open weight.

On HumanEval, Claude 2 leads at 71.2% vs DeepSeek V2 Lite at 57.3%. On MMLU, Claude 2 leads at 78.5% vs DeepSeek V2 Lite at 55.7%. On GSM8K, Claude 2 leads at 88% vs DeepSeek V2 Lite at 72%.