Compare AI models

Specifications
Parameters
16B1B/4B/12B/27B
Context window
32k—
API pricing
Cheapest input
—$0.04DeepInfra
Cheapest output
—$0.08DeepInfra
—$15.00DeepInfra · bf16
Benchmarks
HumanEval
57.3%87.8%
MATH
27.9%89%
MMLU
55.7%76.9%
GSM8K
72%95.9%
Overview
CompanyDeepSeekGoogle
Release dateMay 16 2024Mar 12 2025
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

Gemma 3 leads DeepSeek V2 Lite on 4 of the 4 benchmarks they both report (HumanEval, MATH, MMLU, GSM8K). DeepSeek V2 Lite shipped 300 days before Gemma 3, so benchmark comparisons should account for the intervening progress.

DeepSeek V2 Lite has 16B parameters, while Gemma 3 has 1B/4B/12B/27B.

On HumanEval, Gemma 3 leads at 87.8% vs DeepSeek V2 Lite at 57.3%. On MATH, Gemma 3 leads at 89% vs DeepSeek V2 Lite at 27.9%. On MMLU, Gemma 3 leads at 76.9% vs DeepSeek V2 Lite at 55.7%. On GSM8K, Gemma 3 leads at 95.9% vs DeepSeek V2 Lite at 72%.