benchlm.ai

DeepSeek V3.1 (Reasoning) vs Grok 3 [Beta]

Side-by-side benchmark comparison across knowledge, coding, math, and reasoning.

Quick Verdict

Grok 3 [Beta] wins overall with a score of 33 vs 25 (8 point difference).Grok 3 [Beta] wins 4 out of 4 categories.

DeepSeek V3.1 (Reasoning)

CreatorDeepSeek
TypeOpen Weight
Context128K
Score25

Grok 3 [Beta]

CreatorxAI
TypeProprietary
Context128K
Score33

Knowledge

Grok 3 [Beta]

DeepSeek V3.1 (Reasoning)

31.8

Grok 3 [Beta]

39.8

34
MMLU
42
33
GPQA
41
31
SuperGPQA
39
29
OpenBookQA
37

Coding

Grok 3 [Beta]

DeepSeek V3.1 (Reasoning)

26

Grok 3 [Beta]

34

26
HumanEval
34

Mathematics

Grok 3 [Beta]

DeepSeek V3.1 (Reasoning)

33

Grok 3 [Beta]

41

34
AIME 2023
42
36
AIME 2024
44
35
AIME 2025
43
30
HMMT Feb 2023
38
32
HMMT Feb 2024
40
31
HMMT Feb 2025
39
33
BRUMO 2025
41

Reasoning

Grok 3 [Beta]

DeepSeek V3.1 (Reasoning)

31

Grok 3 [Beta]

39

32
SimpleQA
40
30
MuSR
38

Frequently Asked Questions

Which is better, DeepSeek V3.1 (Reasoning) or Grok 3 [Beta]?

Grok 3 [Beta] scores higher overall with 33 vs 25, a difference of 8 points across all benchmarks.

Which is better for knowledge tasks, DeepSeek V3.1 (Reasoning) or Grok 3 [Beta]?

Grok 3 [Beta] leads in knowledge tasks with an average score of 39.8 vs 31.8.

Which is better for coding, DeepSeek V3.1 (Reasoning) or Grok 3 [Beta]?

Grok 3 [Beta] leads in coding with an average score of 34 vs 26.

Which is better for math, DeepSeek V3.1 (Reasoning) or Grok 3 [Beta]?

Grok 3 [Beta] leads in math with an average score of 41 vs 33.

Which is better for reasoning, DeepSeek V3.1 (Reasoning) or Grok 3 [Beta]?

Grok 3 [Beta] leads in reasoning with an average score of 39 vs 31.

Related Comparisons

DeepSeek V3.1 (Reasoning) vs GPT-5.4DeepSeek V3.1 (Reasoning) vs Gemini 3.1 ProDeepSeek V3.1 (Reasoning) vs Claude Opus 4.6DeepSeek V3.1 (Reasoning) vs GPT-5.3 CodexDeepSeek V3.1 (Reasoning) vs Grok 4.1DeepSeek V3.1 (Reasoning) vs GPT-5.2