GeekRadar / The comparison desk
COMPARE.
AI models
anthropic

Claude Sonnet 4

AI model Released

Context window
200,000 tokens
Maximum output
64,000 tokens
Open weights
No

Capability & preference

General capability · ECI

A cross-benchmark capability estimate. Index points, not a percentage; intervals show uncertainty.

Claude Sonnet 4 141.6* Claude Sonnet 4 · 95% interval 138.8–142.9

Human preference · Arena

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

Claude Sonnet 4 1339.2 claude-sonnet-4-20250514 · 95% interval 1334.8–1343.7 · 39,163 votes · 2026-10-15

Model performance

Radar task score: four tests, equal weight. An experimental summary of these tasks, not a universal intelligence rating.

Claude Sonnet 4

Radar task score

Unrated

1/4 core tests · not enough evidence

1 benchmark results for this configuration

Scientific reasoning

GPQA Diamond · 25%

Expert-written science questions; strict multiple-choice scoring. Chance accuracy is 25%.

Claude Sonnet 4 77.8*% Approx. 95% interval: 72.6–83.1% · Older result

Novel problem solving

Mystery Game Puzzles · 25%

Best-next-move puzzles in an undisclosed game variant. A narrow reasoning test, not general intelligence.

Claude Sonnet 4 Not available

Advanced mathematics

FrontierMath Tiers 1-3 · 25%

Private v2 mathematics problems with Python tools. Earlier versions and Tier 4 are excluded.

Claude Sonnet 4 Not available

Factual answers

SimpleQA Verified · 25%

Correct factual answers under a guessing prompt, graded by a model. Not a hallucination or abstention score.

Claude Sonnet 4 Not available

Bars show reported task scores. Whiskers are approximate 95% intervals where standard errors are available. Different reasoning budgets are not equal-cost tests. Missing results are never zero.

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

Claude Sonnet 4

1,350.03 rating

Confidence interval: 1,337.68 – 1,362.38

2376 votes

Category
chinese
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference

1,340.69 rating

Confidence interval: 1,329.06 – 1,352.32

2842 votes

Category
chinese
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference
Arena text preference / codinghuman preference · 2026-10-15 · rating

Claude Sonnet 4

1,414.15 rating

Confidence interval: 1,406.26 – 1,422.04

6348 votes

Category
coding
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference

1,379.18 rating

Confidence interval: 1,371.79 – 1,386.58

7307 votes

Category
coding
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference
Arena text preference / creative_writinghuman preference · 2026-10-15 · rating

Claude Sonnet 4

1,344.99 rating

Confidence interval: 1,335.50 – 1,354.47

4248 votes

Category
creative_writing
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference

1,342.18 rating

Confidence interval: 1,333.28 – 1,351.09

4905 votes

Category
creative_writing
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference
Arena text preference / englishhuman preference · 2026-10-15 · rating

Claude Sonnet 4

1,364.41 rating

Confidence interval: 1,358.61 – 1,370.22

14847 votes

Category
english
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference

1,343.93 rating

Confidence interval: 1,338.30 – 1,349.56

17205 votes

Category
english
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference
Arena text preference / exclude_tieshuman preference · 2026-10-15 · rating

Claude Sonnet 4

1,311.07 rating

Confidence interval: 1,305.19 – 1,316.95

25040 votes

Category
exclude_ties
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference

1,295.04 rating

Confidence interval: 1,289.24 – 1,300.83

28586 votes

Category
exclude_ties
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference
Arena text preference / experthuman preference · 2026-10-15 · rating

Claude Sonnet 4

1,372.22 rating

Confidence interval: 1,357.75 – 1,386.69

1702 votes

Category
expert
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference

1,335.31 rating

Confidence interval: 1,322.10 – 1,348.52

2091 votes

Category
expert
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference
Arena text preference / frenchhuman preference · 2026-10-15 · rating

Claude Sonnet 4

1,363.06 rating

Confidence interval: 1,337.82 – 1,388.30

570 votes

Category
french
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference

1,355.67 rating

Confidence interval: 1,328.85 – 1,382.50

472 votes

Category
french
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference
Arena text preference / germanhuman preference · 2026-10-15 · rating

Claude Sonnet 4

1,330.94 rating

Confidence interval: 1,312.20 – 1,349.68

1081 votes

Category
german
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference

1,317.39 rating

Confidence interval: 1,296.51 – 1,338.27

860 votes

Category
german
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference
Arena text preference / hard_promptshuman preference · 2026-10-15 · rating

Claude Sonnet 4

1,372.43 rating

Confidence interval: 1,366.70 – 1,378.15

15413 votes

Category
hard_prompts
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference

1,354.86 rating

Confidence interval: 1,349.27 – 1,360.45

17418 votes

Category
hard_prompts
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference
Arena text preference / hard_prompts_englishhuman preference · 2026-10-15 · rating

Claude Sonnet 4

1,384.99 rating

Confidence interval: 1,377.36 – 1,392.61

6982 votes

Category
hard_prompts_english
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference

1,358.55 rating

Confidence interval: 1,351.28 – 1,365.81

7899 votes

Category
hard_prompts_english
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference
Arena text preference / industry_business_and_management_and_financial_operationshuman preference · 2026-10-15 · rating

Claude Sonnet 4

1,318.44 rating

Confidence interval: 1,310.44 – 1,326.44

6541 votes

Category
industry_business_and_management_and_financial_operations
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference

1,314.20 rating

Confidence interval: 1,305.82 – 1,322.58

5771 votes

Category
industry_business_and_management_and_financial_operations
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference
Arena text preference / industry_entertainment_and_sports_and_mediahuman preference · 2026-10-15 · rating

Claude Sonnet 4

1,331.88 rating

Confidence interval: 1,323.83 – 1,339.94

6040 votes

Category
industry_entertainment_and_sports_and_media
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference

1,320.30 rating

Confidence interval: 1,312.52 – 1,328.07

6850 votes

Category
industry_entertainment_and_sports_and_media
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference
Arena text preference / industry_legal_and_governmenthuman preference · 2026-10-15 · rating

Claude Sonnet 4

1,341.45 rating

Confidence interval: 1,329.33 – 1,353.58

2622 votes

Category
industry_legal_and_government
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference

1,336.60 rating

Confidence interval: 1,323.44 – 1,349.77

2207 votes

Category
industry_legal_and_government
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference
Arena text preference / industry_life_and_physical_and_social_sciencehuman preference · 2026-10-15 · rating

Claude Sonnet 4

1,345.69 rating

Confidence interval: 1,337.34 – 1,354.04

5611 votes

Category
industry_life_and_physical_and_social_science
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference

1,339.93 rating

Confidence interval: 1,331.85 – 1,348.01

6379 votes

Category
industry_life_and_physical_and_social_science
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference
Arena text preference / industry_mathematicalhuman preference · 2026-10-15 · rating

Claude Sonnet 4

1,379.22 rating

Confidence interval: 1,365.88 – 1,392.57

1915 votes

Category
industry_mathematical
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference

1,357.73 rating

Confidence interval: 1,345.40 – 1,370.06

2322 votes

Category
industry_mathematical
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference
Arena text preference / industry_medicine_and_healthcarehuman preference · 2026-10-15 · rating

Claude Sonnet 4

1,330.51 rating

Confidence interval: 1,317.51 – 1,343.51

2364 votes

Category
industry_medicine_and_healthcare
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference

1,326.35 rating

Confidence interval: 1,312.98 – 1,339.73

2123 votes

Category
industry_medicine_and_healthcare
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference
Arena text preference / industry_software_and_it_serviceshuman preference · 2026-10-15 · rating

Claude Sonnet 4

1,383.69 rating

Confidence interval: 1,377.33 – 1,390.04

11254 votes

Category
industry_software_and_it_services
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference

1,360.64 rating

Confidence interval: 1,354.55 – 1,366.73

13140 votes

Category
industry_software_and_it_services
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference
Arena text preference / industry_writing_and_literature_and_languagehuman preference · 2026-10-15 · rating

Claude Sonnet 4

1,356.91 rating

Confidence interval: 1,349.50 – 1,364.31

7373 votes

Category
industry_writing_and_literature_and_language
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference

1,350.85 rating

Confidence interval: 1,343.76 – 1,357.93

8500 votes

Category
industry_writing_and_literature_and_language
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference
Arena text preference / instruction_followinghuman preference · 2026-10-15 · rating

Claude Sonnet 4

1,376.05 rating

Confidence interval: 1,369.03 – 1,383.07

8523 votes

Category
instruction_following
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference

1,351.99 rating

Confidence interval: 1,345.34 – 1,358.65

9910 votes

Category
instruction_following
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference
Arena text preference / japanesehuman preference · 2026-10-15 · rating

Claude Sonnet 4

1,301.65 rating

Confidence interval: 1,281.73 – 1,321.57

969 votes

Category
japanese
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference

1,287.13 rating

Confidence interval: 1,264.84 – 1,309.41

821 votes

Category
japanese
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference
Arena text preference / koreanhuman preference · 2026-10-15 · rating

Claude Sonnet 4

1,290.73 rating

Confidence interval: 1,270.17 – 1,311.29

930 votes

Category
korean
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference

1,271.35 rating

Confidence interval: 1,248.47 – 1,294.24

731 votes

Category
korean
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference
Arena text preference / longer_queryhuman preference · 2026-10-15 · rating

Claude Sonnet 4

1,397.78 rating

Confidence interval: 1,390.10 – 1,405.47

6721 votes

Category
longer_query
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference

1,379.46 rating

Confidence interval: 1,372.18 – 1,386.74

7699 votes

Category
longer_query
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference
Arena text preference / mathhuman preference · 2026-10-15 · rating

Claude Sonnet 4

1,375.07 rating

Confidence interval: 1,362.20 – 1,387.93

2015 votes

Category
math
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference

1,358.99 rating

Confidence interval: 1,347.28 – 1,370.69

2449 votes

Category
math
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference
Arena text preference / multi_turnhuman preference · 2026-10-15 · rating

Claude Sonnet 4

1,376.44 rating

Confidence interval: 1,368.08 – 1,384.79

5633 votes

Category
multi_turn
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference

1,364.73 rating

Confidence interval: 1,356.81 – 1,372.64

6816 votes

Category
multi_turn
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference
Arena text preference / non_englishhuman preference · 2026-10-15 · rating

Claude Sonnet 4

1,332.63 rating

Confidence interval: 1,327.26 – 1,338.00

19280 votes

Category
non_english
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference

1,330.07 rating

Confidence interval: 1,324.76 – 1,335.38

21949 votes

Category
non_english
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference
Arena text preference / overallhuman preference · 2026-10-15 · rating

Claude Sonnet 4

1,350.88 rating

Confidence interval: 1,346.39 – 1,355.38

34132 votes

Category
overall
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference

1,339.25 rating

Confidence interval: 1,334.82 – 1,343.67

39163 votes

Category
overall
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference
Arena text preference / polishhuman preference · 2026-10-15 · rating

Claude Sonnet 4

1,324.35 rating

Confidence interval: 1,312.98 – 1,335.71

3272 votes

Category
polish
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference

1,321.74 rating

Confidence interval: 1,309.81 – 1,333.68

2915 votes

Category
polish
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference
Arena text preference / russianhuman preference · 2026-10-15 · rating

Claude Sonnet 4

1,355.35 rating

Confidence interval: 1,343.50 – 1,367.19

2411 votes

Category
russian
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference

1,352.71 rating

Confidence interval: 1,342.18 – 1,363.25

3214 votes

Category
russian
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference
Arena text preference / spanishhuman preference · 2026-10-15 · rating

Claude Sonnet 4

1,356.67 rating

Confidence interval: 1,335.33 – 1,378.01

835 votes

Category
spanish
Evaluated configuration
claude-sonnet-4-20250514-thinking-32k
Evaluation type
human preference

1,349.42 rating

Confidence interval: 1,328.86 – 1,369.98

891 votes

Category
spanish
Evaluated configuration
claude-sonnet-4-20250514
Evaluation type
human preference

Run it locally

Inference · memory estimate

Claude Sonnet 4

Closed weights. No supported local deployment or GPU requirement is published.

API cost

USD / month

Standard text-token estimate. Excludes taxes, caching, batch discounts, reasoning and multimodal surcharges. Provider limits and regional availability may differ.

Claude Sonnet 4

Capabilities

Context window
200,000 tokens
Maximum output
64,000 tokens
Reasoning
Yes
Tool calling
Yes
Open weights
No
Knowledge cutoff
2025-03-31
Model license
Proprietary

Input modalities

  • text
  • image
  • pdf

Output modalities

  • text

* Provisional data