GeekRadar / The comparison desk

COMPARE.

AI models

stepfun

Step-3

Model evaluation

Capability & preference

Human preference · Arena

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

Step-3 1350.4 Step-3 · 95% interval 1342.9–1357.8 · 6,291 votes · 2026-10-28

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 · 2026-10-28 · rating

Step-3

1,397.36 rating

Confidence interval: 1,365.10 – 1,429.63

313 votes

Oct 28, 2026

Category: chinese
Evaluated configuration: Step-3
Evaluation type: human preference

Arena text preference / codinghuman preference · 2026-10-28 · rating

Step-3

1,366.91 rating

Confidence interval: 1,350.39 – 1,383.42

1213 votes

Oct 28, 2026

Category: coding
Evaluated configuration: Step-3
Evaluation type: human preference

Arena text preference / creative_writinghuman preference · 2026-10-28 · rating

Step-3

1,321.40 rating

Confidence interval: 1,300.57 – 1,342.22

770 votes

Oct 28, 2026

Category: creative_writing
Evaluated configuration: Step-3
Evaluation type: human preference

Arena text preference / englishhuman preference · 2026-10-28 · rating

Step-3

1,372.98 rating

Confidence interval: 1,362.03 – 1,383.93

2775 votes

Oct 28, 2026

Category: english
Evaluated configuration: Step-3
Evaluation type: human preference

Arena text preference / exclude_tieshuman preference · 2026-10-28 · rating

Step-3

1,311.23 rating

Confidence interval: 1,300.67 – 1,321.80

4454 votes

Oct 28, 2026

Category: exclude_ties
Evaluated configuration: Step-3
Evaluation type: human preference

Arena text preference / experthuman preference · 2026-10-28 · rating

Step-3

1,332.27 rating

Confidence interval: 1,297.50 – 1,367.05

264 votes

Oct 28, 2026

Category: expert
Evaluated configuration: Step-3
Evaluation type: human preference

Arena text preference / germanhuman preference · 2026-10-28 · rating

Step-3

1,370.68 rating

Confidence interval: 1,324.77 – 1,416.59

153 votes

Oct 28, 2026

Category: german
Evaluated configuration: Step-3
Evaluation type: human preference

Arena text preference / hard_promptshuman preference · 2026-10-28 · rating

Step-3

1,354.75 rating

Confidence interval: 1,344.10 – 1,365.40

2961 votes

Oct 28, 2026

Category: hard_prompts
Evaluated configuration: Step-3
Evaluation type: human preference

Arena text preference / hard_prompts_englishhuman preference · 2026-10-28 · rating

Step-3

1,378.23 rating

Confidence interval: 1,362.35 – 1,394.12

1327 votes

Oct 28, 2026

Category: hard_prompts_english
Evaluated configuration: Step-3
Evaluation type: human preference

Arena text preference / industry_business_and_management_and_financial_operationshuman preference · 2026-10-28 · rating

Step-3

1,342.56 rating

Confidence interval: 1,325.77 – 1,359.35

1142 votes

Oct 28, 2026

Category: industry_business_and_management_and_financial_operations
Evaluated configuration: Step-3
Evaluation type: human preference

Arena text preference / industry_entertainment_and_sports_and_mediahuman preference · 2026-10-28 · rating

Step-3

1,331.29 rating

Confidence interval: 1,314.11 – 1,348.47

1088 votes

Oct 28, 2026

Category: industry_entertainment_and_sports_and_media
Evaluated configuration: Step-3
Evaluation type: human preference

Arena text preference / industry_legal_and_governmenthuman preference · 2026-10-28 · rating

Step-3

1,336.95 rating

Confidence interval: 1,308.38 – 1,365.51

407 votes

Oct 28, 2026

Category: industry_legal_and_government
Evaluated configuration: Step-3
Evaluation type: human preference

Arena text preference / industry_life_and_physical_and_social_sciencehuman preference · 2026-10-28 · rating

Step-3

1,353.71 rating

Confidence interval: 1,335.51 – 1,371.90

1068 votes

Oct 28, 2026

Category: industry_life_and_physical_and_social_science
Evaluated configuration: Step-3
Evaluation type: human preference

Arena text preference / industry_mathematicalhuman preference · 2026-10-28 · rating

Step-3

1,375.24 rating

Confidence interval: 1,341.92 – 1,408.55

299 votes

Oct 28, 2026

Category: industry_mathematical
Evaluated configuration: Step-3
Evaluation type: human preference

Arena text preference / industry_medicine_and_healthcarehuman preference · 2026-10-28 · rating

Step-3

1,361.56 rating

Confidence interval: 1,331.13 – 1,391.98

349 votes

Oct 28, 2026

Category: industry_medicine_and_healthcare
Evaluated configuration: Step-3
Evaluation type: human preference

Arena text preference / industry_software_and_it_serviceshuman preference · 2026-10-28 · rating

Step-3

1,364.97 rating

Confidence interval: 1,352.33 – 1,377.61

2090 votes

Oct 28, 2026

Category: industry_software_and_it_services
Evaluated configuration: Step-3
Evaluation type: human preference

Arena text preference / industry_writing_and_literature_and_languagehuman preference · 2026-10-28 · rating

Step-3

1,326.18 rating

Confidence interval: 1,310.97 – 1,341.39

1386 votes

Oct 28, 2026

Category: industry_writing_and_literature_and_language
Evaluated configuration: Step-3
Evaluation type: human preference

Arena text preference / instruction_followinghuman preference · 2026-10-28 · rating

Step-3

1,332.23 rating

Confidence interval: 1,317.85 – 1,346.61

1636 votes

Oct 28, 2026

Category: instruction_following
Evaluated configuration: Step-3
Evaluation type: human preference

Arena text preference / koreanhuman preference · 2026-10-28 · rating

Step-3

1,269.26 rating

Confidence interval: 1,224.42 – 1,314.10

181 votes

Oct 28, 2026

Category: korean
Evaluated configuration: Step-3
Evaluation type: human preference

Arena text preference / longer_queryhuman preference · 2026-10-28 · rating

Step-3

1,325.52 rating

Confidence interval: 1,309.34 – 1,341.70

1274 votes

Oct 28, 2026

Category: longer_query
Evaluated configuration: Step-3
Evaluation type: human preference

Arena text preference / mathhuman preference · 2026-10-28 · rating

Step-3

1,366.58 rating

Confidence interval: 1,335.35 – 1,397.81

348 votes

Oct 28, 2026

Category: math
Evaluated configuration: Step-3
Evaluation type: human preference

Arena text preference / multi_turnhuman preference · 2026-10-28 · rating

Step-3

1,340.17 rating

Confidence interval: 1,321.94 – 1,358.39

1036 votes

Oct 28, 2026

Category: multi_turn
Evaluated configuration: Step-3
Evaluation type: human preference

Arena text preference / non_englishhuman preference · 2026-10-28 · rating

Step-3

1,327.32 rating

Confidence interval: 1,317.52 – 1,337.13

3513 votes

Oct 28, 2026

Category: non_english
Evaluated configuration: Step-3
Evaluation type: human preference

Arena text preference / overallhuman preference · 2026-10-28 · rating

Step-3

1,350.36 rating

Confidence interval: 1,342.92 – 1,357.80

6291 votes

Oct 28, 2026

Category: overall
Evaluated configuration: Step-3
Evaluation type: human preference

Arena text preference / polishhuman preference · 2026-10-28 · rating

Step-3

1,300.55 rating

Confidence interval: 1,272.92 – 1,328.17

449 votes

Oct 28, 2026

Category: polish
Evaluated configuration: Step-3
Evaluation type: human preference

Arena text preference / russianhuman preference · 2026-10-28 · rating

Step-3

1,331.33 rating

Confidence interval: 1,305.76 – 1,356.89

489 votes

Oct 28, 2026

Category: russian
Evaluated configuration: Step-3
Evaluation type: human preference

Arena text preference / spanishhuman preference · 2026-10-28 · rating

Step-3

1,370.94 rating

Confidence interval: 1,329.58 – 1,412.29

211 votes

Oct 28, 2026

Category: spanish
Evaluated configuration: Step-3
Evaluation type: human preference

Run it locally

Inference · memory estimate

Step-3

Not enough verified architecture information for a reliable VRAM estimate.

Capabilities

Model license: Apache 2.0