Compare AI models

Specifications
Parameters
16B2B/7B
Context window
32k8k
Benchmarks
HumanEval
57.3%32.3%
MATH
27.9%24.3%
MMLU
55.7%64.3%
GSM8K
72%46.4%
Overview
CompanyDeepSeekGoogle
Release dateMay 16 2024Feb 21 2024
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

DeepSeek V2 Lite leads Gemma 1 on 3 of the 4 benchmarks they both report (HumanEval, MATH, MMLU, GSM8K). Gemma 1 shipped 85 days before DeepSeek V2 Lite, so benchmark comparisons should account for the intervening progress.

DeepSeek V2 Lite has 16B parameters, while Gemma 1 has 2B/7B. Context windows are 32k (DeepSeek V2 Lite) vs 8k (Gemma 1).

On HumanEval, DeepSeek V2 Lite leads at 57.3% vs Gemma 1 at 32.3%. On MATH, DeepSeek V2 Lite leads at 27.9% vs Gemma 1 at 24.3%. On MMLU, Gemma 1 leads at 64.3% vs DeepSeek V2 Lite at 55.7%. On GSM8K, DeepSeek V2 Lite leads at 72% vs Gemma 1 at 46.4%.