Compare AI models

Specifications
Parameters
236B—
Context window
128k200k
Benchmarks
HumanEval
84.76%99.3%
Overview
CompanyDeepSeekOpenAI
Release dateJun 28 2024Apr 16 2025
AccessOpen WeightClosed
Model detailsView modelView model

Frequently asked questions

o4-mini-high leads DeepSeek V2-0628 on 1 of the 1 benchmark they both report (HumanEval). DeepSeek V2-0628 shipped 292 days before o4-mini-high, so benchmark comparisons should account for the intervening progress.

Context windows are 128k (DeepSeek V2-0628) vs 200k (o4-mini-high). DeepSeek V2-0628 is open weight, while o4-mini-high is closed.

On HumanEval, o4-mini-high leads at 99.3% vs DeepSeek V2-0628 at 84.76%.