GeekRadar / The comparison desk
COMPARE.
AI models
deepseek

deepseek-v4-pro-high-preview

Model evaluation

Capability & preference

Human preference · Arena

Overall text preferences from head-to-head votes. Not factual accuracy or a task-success rate.

deepseek-v4-pro-high-preview 1444.9 deepseek-v4-pro-high-preview · 95% interval 1440.9–1448.8 · 54,902 votes · 2026-10-15

Recorded results

Only the same benchmark, version, workload, scenario and unit share a row. Missing results are not zero. System results include the full submitted hardware and software configuration.

Arena text preference / chinesehuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,485.85 rating

Confidence interval: 1,474.99 – 1,496.70

3345 votes

Category
chinese
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference
Arena text preference / codinghuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,452.99 rating

Confidence interval: 1,446.94 – 1,459.04

16007 votes

Category
coding
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference
Arena text preference / creative_writinghuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,440.89 rating

Confidence interval: 1,433.60 – 1,448.18

9383 votes

Category
creative_writing
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference
Arena text preference / englishhuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,454.88 rating

Confidence interval: 1,449.73 – 1,460.04

23991 votes

Category
english
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference
Arena text preference / exclude_tieshuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,440.69 rating

Confidence interval: 1,435.62 – 1,445.75

41747 votes

Category
exclude_ties
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference
Arena text preference / experthuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,449.35 rating

Confidence interval: 1,440.78 – 1,457.92

5541 votes

Category
expert
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference
Arena text preference / frenchhuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,450.94 rating

Confidence interval: 1,436.32 – 1,465.55

2067 votes

Category
french
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference
Arena text preference / germanhuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,447.78 rating

Confidence interval: 1,427.91 – 1,467.66

928 votes

Category
german
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference
Arena text preference / hard_promptshuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,452.23 rating

Confidence interval: 1,447.60 – 1,456.85

36937 votes

Category
hard_prompts
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference
Arena text preference / hard_prompts_englishhuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,460.31 rating

Confidence interval: 1,454.39 – 1,466.23

16891 votes

Category
hard_prompts_english
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference
Arena text preference / industry_business_and_management_and_financial_operationshuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,427.66 rating

Confidence interval: 1,420.92 – 1,434.41

11042 votes

Category
industry_business_and_management_and_financial_operations
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference
Arena text preference / industry_entertainment_and_sports_and_mediahuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,427.14 rating

Confidence interval: 1,420.52 – 1,433.76

12393 votes

Category
industry_entertainment_and_sports_and_media
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference
Arena text preference / industry_legal_and_governmenthuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,460.14 rating

Confidence interval: 1,450.56 – 1,469.71

4336 votes

Category
industry_legal_and_government
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference
Arena text preference / industry_life_and_physical_and_social_sciencehuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,462.62 rating

Confidence interval: 1,455.49 – 1,469.75

8797 votes

Category
industry_life_and_physical_and_social_science
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference
Arena text preference / industry_mathematicalhuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,453.94 rating

Confidence interval: 1,442.54 – 1,465.34

2991 votes

Category
industry_mathematical
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference
Arena text preference / industry_medicine_and_healthcarehuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,468.73 rating

Confidence interval: 1,458.62 – 1,478.84

3852 votes

Category
industry_medicine_and_healthcare
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference
Arena text preference / industry_software_and_it_serviceshuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,456.56 rating

Confidence interval: 1,451.22 – 1,461.90

22427 votes

Category
industry_software_and_it_services
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference
Arena text preference / industry_writing_and_literature_and_languagehuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,437.59 rating

Confidence interval: 1,431.24 – 1,443.94

13733 votes

Category
industry_writing_and_literature_and_language
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference
Arena text preference / instruction_followinghuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,435.84 rating

Confidence interval: 1,430.19 – 1,441.50

19061 votes

Category
instruction_following
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference
Arena text preference / japanesehuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,414.17 rating

Confidence interval: 1,390.19 – 1,438.15

686 votes

Category
japanese
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference
Arena text preference / koreanhuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,425.43 rating

Confidence interval: 1,405.72 – 1,445.14

979 votes

Category
korean
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference
Arena text preference / longer_queryhuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,446.12 rating

Confidence interval: 1,440.73 – 1,451.50

24726 votes

Category
longer_query
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference
Arena text preference / mathhuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,454.38 rating

Confidence interval: 1,442.63 – 1,466.13

2714 votes

Category
math
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference
Arena text preference / multi_turnhuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,445.36 rating

Confidence interval: 1,438.23 – 1,452.49

9693 votes

Category
multi_turn
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference
Arena text preference / non_englishhuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,431.29 rating

Confidence interval: 1,426.52 – 1,436.05

30907 votes

Category
non_english
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference
Arena text preference / overallhuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,444.89 rating

Confidence interval: 1,440.95 – 1,448.84

54902 votes

Category
overall
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference
Arena text preference / polishhuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,453.80 rating

Confidence interval: 1,435.68 – 1,471.92

1093 votes

Category
polish
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference
Arena text preference / russianhuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,448.99 rating

Confidence interval: 1,440.74 – 1,457.25

6149 votes

Category
russian
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference
Arena text preference / spanishhuman preference · 2026-10-15 · rating

deepseek-v4-pro-high-preview

1,448.80 rating

Confidence interval: 1,433.42 – 1,464.18

1682 votes

Category
spanish
Evaluated configuration
deepseek-v4-pro-high-preview
Evaluation type
human preference

Run it locally

Inference · memory estimate

deepseek-v4-pro-high-preview

Not enough verified architecture information for a reliable VRAM estimate.

Capabilities

Model license
MIT