Refresh Models —Deals —Regulatory —Sources —

Questions this page answers

Which is better, DeepSeek-V2 or Mistral 7B?: Across 12 shared benchmarks, DeepSeek-V2 scores higher on 12 and Mistral 7B on 0. The widest gap is MATH, where DeepSeek-V2 scores 43.6 against 13.1.

DeepSeek-V2 vs Mistral 7B

Across 12 shared benchmarks, DeepSeek-V2 scores higher on 12 and Mistral 7B on 0. The widest gap is MATH, where DeepSeek-V2 scores 43.6 against 13.1.

DeepSeekvsMistral12 shared benchmarks12–0 head-to-head

Benchmark	DeepSeek-V2	Mistral 7B
arc_challenge	92.4	60
bbh	78.9	56.1
C-Eval	81.7	47.4
GPQA Diamond	35.3	24.7
GSM8K	79.2	52.2
hellaswag	87.8	83.2
humaneval	48.8	29.3
MATH	43.6	13.1
MBPP	66.6	51.1
mmlu	78.5	64.2
MMLU-Pro	58.5	30.9
winogrande	86.3	78.4

Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing.