Compare AI models

Specifications
Parameters
236B24B
Context window
128k32k
API pricing
Cheapest input
—$0.04DeepInfra
Cheapest output
—$0.064DeepInfra
—$14.00DeepInfra · fp8
Benchmarks
HumanEval
84.76%84.8%
MATH
71.02%70.6%
Overview
CompanyDeepSeekMistral
Release dateJun 28 2024Jan 30 2025
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

DeepSeek V2-0628 and Mistral Small 3 are evenly matched across the 2 benchmarks they both report (HumanEval, MATH). DeepSeek V2-0628 shipped 216 days before Mistral Small 3, so benchmark comparisons should account for the intervening progress.

DeepSeek V2-0628 has 236B parameters, while Mistral Small 3 has 24B. Context windows are 128k (DeepSeek V2-0628) vs 32k (Mistral Small 3).

On HumanEval, Mistral Small 3 leads at 84.8% vs DeepSeek V2-0628 at 84.76%. On MATH, DeepSeek V2-0628 leads at 71.02% vs Mistral Small 3 at 70.6%.