GeekRadar / The comparison desk
COMPARE.
AI models
google

gemma-4-31B

AI model

Context window
262,144 tokens
Open weights
Yes

Capability & preference

Human preference · Arena

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

gemma-4-31B 1443.0 Gemma 4 31B · 95% interval 1435.6–1450.4 · 6,138 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

gemma-4-31B

1,475.38 rating

Confidence interval: 1,450.44 – 1,500.32

522 votes

Category
chinese
Evaluated configuration
Gemma 4 31B
Evaluation type
human preference
Arena text preference / codinghuman preference · 2026-10-28 · rating

gemma-4-31B

1,459.13 rating

Confidence interval: 1,444.11 – 1,474.16

1411 votes

Category
coding
Evaluated configuration
Gemma 4 31B
Evaluation type
human preference
Arena text preference / creative_writinghuman preference · 2026-10-28 · rating

gemma-4-31B

1,416.24 rating

Confidence interval: 1,397.58 – 1,434.89

971 votes

Category
creative_writing
Evaluated configuration
Gemma 4 31B
Evaluation type
human preference
Arena text preference / englishhuman preference · 2026-10-28 · rating

gemma-4-31B

1,450.98 rating

Confidence interval: 1,439.17 – 1,462.78

2329 votes

Category
english
Evaluated configuration
Gemma 4 31B
Evaluation type
human preference
Arena text preference / exclude_tieshuman preference · 2026-10-28 · rating

gemma-4-31B

1,441.56 rating

Confidence interval: 1,430.82 – 1,452.29

4222 votes

Category
exclude_ties
Evaluated configuration
Gemma 4 31B
Evaluation type
human preference
Arena text preference / experthuman preference · 2026-10-28 · rating

gemma-4-31B

1,464.63 rating

Confidence interval: 1,439.86 – 1,489.41

502 votes

Category
expert
Evaluated configuration
Gemma 4 31B
Evaluation type
human preference
Arena text preference / frenchhuman preference · 2026-10-28 · rating

gemma-4-31B

1,435.75 rating

Confidence interval: 1,400.83 – 1,470.67

268 votes

Category
french
Evaluated configuration
Gemma 4 31B
Evaluation type
human preference
Arena text preference / hard_promptshuman preference · 2026-10-28 · rating

gemma-4-31B

1,447.66 rating

Confidence interval: 1,438.05 – 1,457.28

3563 votes

Category
hard_prompts
Evaluated configuration
Gemma 4 31B
Evaluation type
human preference
Arena text preference / hard_prompts_englishhuman preference · 2026-10-28 · rating

gemma-4-31B

1,456.37 rating

Confidence interval: 1,440.67 – 1,472.08

1329 votes

Category
hard_prompts_english
Evaluated configuration
Gemma 4 31B
Evaluation type
human preference
Arena text preference / industry_business_and_management_and_financial_operationshuman preference · 2026-10-28 · rating

gemma-4-31B

1,430.21 rating

Confidence interval: 1,413.56 – 1,446.87

1179 votes

Category
industry_business_and_management_and_financial_operations
Evaluated configuration
Gemma 4 31B
Evaluation type
human preference
Arena text preference / industry_entertainment_and_sports_and_mediahuman preference · 2026-10-28 · rating

gemma-4-31B

1,404.76 rating

Confidence interval: 1,387.21 – 1,422.31

1092 votes

Category
industry_entertainment_and_sports_and_media
Evaluated configuration
Gemma 4 31B
Evaluation type
human preference
Arena text preference / industry_legal_and_governmenthuman preference · 2026-10-28 · rating

gemma-4-31B

1,438.07 rating

Confidence interval: 1,412.02 – 1,464.12

452 votes

Category
industry_legal_and_government
Evaluated configuration
Gemma 4 31B
Evaluation type
human preference
Arena text preference / industry_life_and_physical_and_social_sciencehuman preference · 2026-10-28 · rating

gemma-4-31B

1,456.95 rating

Confidence interval: 1,438.16 – 1,475.75

932 votes

Category
industry_life_and_physical_and_social_science
Evaluated configuration
Gemma 4 31B
Evaluation type
human preference
Arena text preference / industry_mathematicalhuman preference · 2026-10-28 · rating

gemma-4-31B

1,465.10 rating

Confidence interval: 1,435.73 – 1,494.48

343 votes

Category
industry_mathematical
Evaluated configuration
Gemma 4 31B
Evaluation type
human preference
Arena text preference / industry_medicine_and_healthcarehuman preference · 2026-10-28 · rating

gemma-4-31B

1,450.44 rating

Confidence interval: 1,422.57 – 1,478.31

422 votes

Category
industry_medicine_and_healthcare
Evaluated configuration
Gemma 4 31B
Evaluation type
human preference
Arena text preference / industry_software_and_it_serviceshuman preference · 2026-10-28 · rating

gemma-4-31B

1,463.30 rating

Confidence interval: 1,451.31 – 1,475.30

2217 votes

Category
industry_software_and_it_services
Evaluated configuration
Gemma 4 31B
Evaluation type
human preference
Arena text preference / industry_writing_and_literature_and_languagehuman preference · 2026-10-28 · rating

gemma-4-31B

1,431.93 rating

Confidence interval: 1,416.66 – 1,447.21

1427 votes

Category
industry_writing_and_literature_and_language
Evaluated configuration
Gemma 4 31B
Evaluation type
human preference
Arena text preference / instruction_followinghuman preference · 2026-10-28 · rating

gemma-4-31B

1,433.59 rating

Confidence interval: 1,419.98 – 1,447.20

1799 votes

Category
instruction_following
Evaluated configuration
Gemma 4 31B
Evaluation type
human preference
Arena text preference / longer_queryhuman preference · 2026-10-28 · rating

gemma-4-31B

1,445.83 rating

Confidence interval: 1,432.42 – 1,459.24

1796 votes

Category
longer_query
Evaluated configuration
Gemma 4 31B
Evaluation type
human preference
Arena text preference / mathhuman preference · 2026-10-28 · rating

gemma-4-31B

1,464.16 rating

Confidence interval: 1,437.53 – 1,490.80

429 votes

Category
math
Evaluated configuration
Gemma 4 31B
Evaluation type
human preference
Arena text preference / multi_turnhuman preference · 2026-10-28 · rating

gemma-4-31B

1,451.87 rating

Confidence interval: 1,434.68 – 1,469.06

1124 votes

Category
multi_turn
Evaluated configuration
Gemma 4 31B
Evaluation type
human preference
Arena text preference / non_englishhuman preference · 2026-10-28 · rating

gemma-4-31B

1,430.80 rating

Confidence interval: 1,421.54 – 1,440.06

3809 votes

Category
non_english
Evaluated configuration
Gemma 4 31B
Evaluation type
human preference
Arena text preference / overallhuman preference · 2026-10-28 · rating

gemma-4-31B

1,442.96 rating

Confidence interval: 1,435.56 – 1,450.36

6138 votes

Category
overall
Evaluated configuration
Gemma 4 31B
Evaluation type
human preference
Arena text preference / polishhuman preference · 2026-10-28 · rating

gemma-4-31B

1,431.28 rating

Confidence interval: 1,393.77 – 1,468.79

203 votes

Category
polish
Evaluated configuration
Gemma 4 31B
Evaluation type
human preference
Arena text preference / russianhuman preference · 2026-10-28 · rating

gemma-4-31B

1,460.76 rating

Confidence interval: 1,441.16 – 1,480.37

825 votes

Category
russian
Evaluated configuration
Gemma 4 31B
Evaluation type
human preference
Arena text preference / spanishhuman preference · 2026-10-28 · rating

gemma-4-31B

1,443.57 rating

Confidence interval: 1,408.18 – 1,478.95

253 votes

Category
spanish
Evaluated configuration
Gemma 4 31B
Evaluation type
human preference
Arena text preference / chinesehuman preference · 2026-10-15 · rating

gemma-4-31B

1,475.33 rating

Confidence interval: 1,450.38 – 1,500.27

522 votes

Category
chinese
Evaluated configuration
gemma-4-31b
Evaluation type
human preference
Arena text preference / codinghuman preference · 2026-10-15 · rating

gemma-4-31B

1,459.22 rating

Confidence interval: 1,444.25 – 1,474.19

1418 votes

Category
coding
Evaluated configuration
gemma-4-31b
Evaluation type
human preference
Arena text preference / creative_writinghuman preference · 2026-10-15 · rating

gemma-4-31B

1,415.44 rating

Confidence interval: 1,396.81 – 1,434.07

974 votes

Category
creative_writing
Evaluated configuration
gemma-4-31b
Evaluation type
human preference
Arena text preference / englishhuman preference · 2026-10-15 · rating

gemma-4-31B

1,451.01 rating

Confidence interval: 1,439.22 – 1,462.81

2332 votes

Category
english
Evaluated configuration
gemma-4-31b
Evaluation type
human preference
Arena text preference / exclude_tieshuman preference · 2026-10-15 · rating

gemma-4-31B

1,441.53 rating

Confidence interval: 1,430.80 – 1,452.26

4227 votes

Category
exclude_ties
Evaluated configuration
gemma-4-31b
Evaluation type
human preference
Arena text preference / experthuman preference · 2026-10-15 · rating

gemma-4-31B

1,464.46 rating

Confidence interval: 1,439.63 – 1,489.29

501 votes

Category
expert
Evaluated configuration
gemma-4-31b
Evaluation type
human preference
Arena text preference / frenchhuman preference · 2026-10-15 · rating

gemma-4-31B

1,435.32 rating

Confidence interval: 1,400.51 – 1,470.13

269 votes

Category
french
Evaluated configuration
gemma-4-31b
Evaluation type
human preference
Arena text preference / hard_promptshuman preference · 2026-10-15 · rating

gemma-4-31B

1,447.75 rating

Confidence interval: 1,438.15 – 1,457.35

3573 votes

Category
hard_prompts
Evaluated configuration
gemma-4-31b
Evaluation type
human preference
Arena text preference / hard_prompts_englishhuman preference · 2026-10-15 · rating

gemma-4-31B

1,456.36 rating

Confidence interval: 1,440.67 – 1,472.04

1332 votes

Category
hard_prompts_english
Evaluated configuration
gemma-4-31b
Evaluation type
human preference
Arena text preference / industry_business_and_management_and_financial_operationshuman preference · 2026-10-15 · rating

gemma-4-31B

1,430.60 rating

Confidence interval: 1,413.95 – 1,447.25

1180 votes

Category
industry_business_and_management_and_financial_operations
Evaluated configuration
gemma-4-31b
Evaluation type
human preference
Arena text preference / industry_entertainment_and_sports_and_mediahuman preference · 2026-10-15 · rating

gemma-4-31B

1,404.10 rating

Confidence interval: 1,386.57 – 1,421.63

1095 votes

Category
industry_entertainment_and_sports_and_media
Evaluated configuration
gemma-4-31b
Evaluation type
human preference
Arena text preference / industry_legal_and_governmenthuman preference · 2026-10-15 · rating

gemma-4-31B

1,438.61 rating

Confidence interval: 1,412.53 – 1,464.70

452 votes

Category
industry_legal_and_government
Evaluated configuration
gemma-4-31b
Evaluation type
human preference
Arena text preference / industry_life_and_physical_and_social_sciencehuman preference · 2026-10-15 · rating

gemma-4-31B

1,457.27 rating

Confidence interval: 1,438.48 – 1,476.05

933 votes

Category
industry_life_and_physical_and_social_science
Evaluated configuration
gemma-4-31b
Evaluation type
human preference
Arena text preference / industry_mathematicalhuman preference · 2026-10-15 · rating

gemma-4-31B

1,464.99 rating

Confidence interval: 1,435.76 – 1,494.21

345 votes

Category
industry_mathematical
Evaluated configuration
gemma-4-31b
Evaluation type
human preference
Arena text preference / industry_medicine_and_healthcarehuman preference · 2026-10-15 · rating

gemma-4-31B

1,450.56 rating

Confidence interval: 1,422.68 – 1,478.44

422 votes

Category
industry_medicine_and_healthcare
Evaluated configuration
gemma-4-31b
Evaluation type
human preference
Arena text preference / industry_software_and_it_serviceshuman preference · 2026-10-15 · rating

gemma-4-31B

1,463.43 rating

Confidence interval: 1,451.46 – 1,475.40

2224 votes

Category
industry_software_and_it_services
Evaluated configuration
gemma-4-31b
Evaluation type
human preference
Arena text preference / industry_writing_and_literature_and_languagehuman preference · 2026-10-15 · rating

gemma-4-31B

1,431.11 rating

Confidence interval: 1,415.85 – 1,446.37

1429 votes

Category
industry_writing_and_literature_and_language
Evaluated configuration
gemma-4-31b
Evaluation type
human preference
Arena text preference / instruction_followinghuman preference · 2026-10-15 · rating

gemma-4-31B

1,433.48 rating

Confidence interval: 1,419.88 – 1,447.07

1803 votes

Category
instruction_following
Evaluated configuration
gemma-4-31b
Evaluation type
human preference
Arena text preference / longer_queryhuman preference · 2026-10-15 · rating

gemma-4-31B

1,446.18 rating

Confidence interval: 1,432.79 – 1,459.58

1801 votes

Category
longer_query
Evaluated configuration
gemma-4-31b
Evaluation type
human preference
Arena text preference / mathhuman preference · 2026-10-15 · rating

gemma-4-31B

1,464.23 rating

Confidence interval: 1,437.65 – 1,490.82

430 votes

Category
math
Evaluated configuration
gemma-4-31b
Evaluation type
human preference
Arena text preference / multi_turnhuman preference · 2026-10-15 · rating

gemma-4-31B

1,452.15 rating

Confidence interval: 1,434.96 – 1,469.35

1124 votes

Category
multi_turn
Evaluated configuration
gemma-4-31b
Evaluation type
human preference
Arena text preference / non_englishhuman preference · 2026-10-15 · rating

gemma-4-31B

1,430.65 rating

Confidence interval: 1,421.40 – 1,439.90

3815 votes

Category
non_english
Evaluated configuration
gemma-4-31b
Evaluation type
human preference
Arena text preference / overallhuman preference · 2026-10-15 · rating

gemma-4-31B

1,442.86 rating

Confidence interval: 1,435.47 – 1,450.26

6147 votes

Category
overall
Evaluated configuration
gemma-4-31b
Evaluation type
human preference
Arena text preference / polishhuman preference · 2026-10-15 · rating

gemma-4-31B

1,431.20 rating

Confidence interval: 1,393.67 – 1,468.74

203 votes

Category
polish
Evaluated configuration
gemma-4-31b
Evaluation type
human preference
Arena text preference / russianhuman preference · 2026-10-15 · rating

gemma-4-31B

1,460.72 rating

Confidence interval: 1,441.13 – 1,480.31

827 votes

Category
russian
Evaluated configuration
gemma-4-31b
Evaluation type
human preference
Arena text preference / spanishhuman preference · 2026-10-15 · rating

gemma-4-31B

1,443.36 rating

Confidence interval: 1,407.96 – 1,478.76

253 votes

Category
spanish
Evaluated configuration
gemma-4-31b
Evaluation type
human preference

Run it locally

Inference · memory estimate

gemma-4-31B

58.3 GiB checkpoint tensor payload · 31.27 B stored tensor elements · BF16

Stored tensors can include buffers and repeated shared weights. A verified unique parameter count is unavailable. Checkpoint size is not the memory required to run the model.

Capabilities

Context window
262,144 tokens
Open weights
Yes
Model license
apache-2.0
Checkpoint tensor payload
58.25 GiB

Input modalities

  • text
  • image

Output modalities

  • text

Processing & platform

Architecture
gemma4