Compare AI models

Specifications
Parameters
16B—
Context window
32k128k
Benchmarks
HumanEval
57.3%88.4%
MATH
27.9%76.1%
MMLU
55.7%87.5%
Overview
CompanyDeepSeekSpaceXSI
Release dateMay 16 2024Aug 13 2024
AccessOpen WeightClosed
Model detailsView modelView model

Frequently asked questions

Grok‑2 leads DeepSeek V2 Lite on 3 of the 3 benchmarks they both report (HumanEval, MATH, MMLU). DeepSeek V2 Lite shipped 89 days before Grok‑2, so benchmark comparisons should account for the intervening progress.

Context windows are 32k (DeepSeek V2 Lite) vs 128k (Grok‑2). DeepSeek V2 Lite is open weight, while Grok‑2 is closed.

On HumanEval, Grok‑2 leads at 88.4% vs DeepSeek V2 Lite at 57.3%. On MATH, Grok‑2 leads at 76.1% vs DeepSeek V2 Lite at 27.9%. On MMLU, Grok‑2 leads at 87.5% vs DeepSeek V2 Lite at 55.7%.