geekradar
Int
Explore GeekRadar

More to explore

Good gear All topics About us Say hello
GeekRadar / The comparison desk
COMPARE.
AI models
deepseek

DeepSeek-V3

AI model Released

API mode
Chat*
Context window
131,072 tokens
Maximum output
8,192 tokens
Open weights
Yes

Capability & preference

General capability · ECI

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

DeepSeek-V3 132.3* DeepSeek-V3 · 95% interval 127.5–135.5

Human preference · Arena

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

DeepSeek-V3 1332.5 DeepSeek-V3 · 95% interval 1327.7–1337.2 · 21,770 votes · 2026-10-28

Model performance

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

DeepSeek-V3

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%.

DeepSeek-V3 56.5*% Approx. 95% interval: 51.1–62.0% · 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.

DeepSeek-V3 Not available

Advanced mathematics

FrontierMath Tiers 1-3 · 25%

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

DeepSeek-V3 Not available

Factual answers

SimpleQA Verified · 25%

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

DeepSeek-V3 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 · 2026-10-28 · rating

DeepSeek-V3

1,338.18 rating

Confidence interval: 1,322.83 – 1,353.53

1450 votes

Category
chinese
Evaluated configuration
DeepSeek-V3
Evaluation type
human preference
Arena text preference / codinghuman preference · 2026-10-28 · rating

DeepSeek-V3

1,324.70 rating

Confidence interval: 1,314.62 – 1,334.79

3280 votes

Category
coding
Evaluated configuration
DeepSeek-V3
Evaluation type
human preference
Arena text preference / creative_writinghuman preference · 2026-10-28 · rating

DeepSeek-V3

1,328.72 rating

Confidence interval: 1,318.70 – 1,338.75

3623 votes

Category
creative_writing
Evaluated configuration
DeepSeek-V3
Evaluation type
human preference
Arena text preference / englishhuman preference · 2026-10-28 · rating

DeepSeek-V3

1,345.76 rating

Confidence interval: 1,339.83 – 1,351.69

12866 votes

Category
english
Evaluated configuration
DeepSeek-V3
Evaluation type
human preference
Arena text preference / exclude_tieshuman preference · 2026-10-28 · rating

DeepSeek-V3

1,288.50 rating

Confidence interval: 1,281.20 – 1,295.80

14106 votes

Category
exclude_ties
Evaluated configuration
DeepSeek-V3
Evaluation type
human preference
Arena text preference / experthuman preference · 2026-10-28 · rating

DeepSeek-V3

1,305.41 rating

Confidence interval: 1,288.89 – 1,321.92

1236 votes

Category
expert
Evaluated configuration
DeepSeek-V3
Evaluation type
human preference
Arena text preference / frenchhuman preference · 2026-10-28 · rating

DeepSeek-V3

1,343.22 rating

Confidence interval: 1,305.70 – 1,380.75

242 votes

Category
french
Evaluated configuration
DeepSeek-V3
Evaluation type
human preference
Arena text preference / germanhuman preference · 2026-10-28 · rating

DeepSeek-V3

1,323.55 rating

Confidence interval: 1,298.35 – 1,348.75

493 votes

Category
german
Evaluated configuration
DeepSeek-V3
Evaluation type
human preference
Arena text preference / hard_promptshuman preference · 2026-10-28 · rating

DeepSeek-V3

1,311.87 rating

Confidence interval: 1,303.83 – 1,319.92

5408 votes

Category
hard_prompts
Evaluated configuration
DeepSeek-V3
Evaluation type
human preference
Arena text preference / hard_prompts_englishhuman preference · 2026-10-28 · rating

DeepSeek-V3

1,323.29 rating

Confidence interval: 1,313.68 – 1,332.89

3637 votes

Category
hard_prompts_english
Evaluated configuration
DeepSeek-V3
Evaluation type
human preference
Arena text preference / industry_business_and_management_and_financial_operationshuman preference · 2026-10-28 · rating

DeepSeek-V3

1,306.91 rating

Confidence interval: 1,294.56 – 1,319.27

2241 votes

Category
industry_business_and_management_and_financial_operations
Evaluated configuration
DeepSeek-V3
Evaluation type
human preference
Arena text preference / industry_entertainment_and_sports_and_mediahuman preference · 2026-10-28 · rating

DeepSeek-V3

1,323.84 rating

Confidence interval: 1,314.57 – 1,333.10

4191 votes

Category
industry_entertainment_and_sports_and_media
Evaluated configuration
DeepSeek-V3
Evaluation type
human preference
Arena text preference / industry_legal_and_governmenthuman preference · 2026-10-28 · rating

DeepSeek-V3

1,341.47 rating

Confidence interval: 1,324.99 – 1,357.94

1312 votes

Category
industry_legal_and_government
Evaluated configuration
DeepSeek-V3
Evaluation type
human preference
Arena text preference / industry_life_and_physical_and_social_sciencehuman preference · 2026-10-28 · rating

DeepSeek-V3

1,332.59 rating

Confidence interval: 1,322.92 – 1,342.26

3766 votes

Category
industry_life_and_physical_and_social_science
Evaluated configuration
DeepSeek-V3
Evaluation type
human preference
Arena text preference / industry_mathematicalhuman preference · 2026-10-28 · rating

DeepSeek-V3

1,313.82 rating

Confidence interval: 1,302.05 – 1,325.59

2326 votes

Category
industry_mathematical
Evaluated configuration
DeepSeek-V3
Evaluation type
human preference
Arena text preference / industry_medicine_and_healthcarehuman preference · 2026-10-28 · rating

DeepSeek-V3

1,345.17 rating

Confidence interval: 1,326.95 – 1,363.40

1060 votes

Category
industry_medicine_and_healthcare
Evaluated configuration
DeepSeek-V3
Evaluation type
human preference
Arena text preference / industry_software_and_it_serviceshuman preference · 2026-10-28 · rating

DeepSeek-V3

1,336.36 rating

Confidence interval: 1,327.96 – 1,344.76

5176 votes

Category
industry_software_and_it_services
Evaluated configuration
DeepSeek-V3
Evaluation type
human preference
Arena text preference / industry_writing_and_literature_and_languagehuman preference · 2026-10-28 · rating

DeepSeek-V3

1,331.10 rating

Confidence interval: 1,322.96 – 1,339.25

5913 votes

Category
industry_writing_and_literature_and_language
Evaluated configuration
DeepSeek-V3
Evaluation type
human preference
Arena text preference / instruction_followinghuman preference · 2026-10-28 · rating

DeepSeek-V3

1,314.89 rating

Confidence interval: 1,308.05 – 1,321.74

8606 votes

Category
instruction_following
Evaluated configuration
DeepSeek-V3
Evaluation type
human preference
Arena text preference / japanesehuman preference · 2026-10-28 · rating

DeepSeek-V3

1,266.42 rating

Confidence interval: 1,242.55 – 1,290.28

534 votes

Category
japanese
Evaluated configuration
DeepSeek-V3
Evaluation type
human preference
Arena text preference / koreanhuman preference · 2026-10-28 · rating

DeepSeek-V3

1,248.42 rating

Confidence interval: 1,213.22 – 1,283.62

295 votes

Category
korean
Evaluated configuration
DeepSeek-V3
Evaluation type
human preference
Arena text preference / longer_queryhuman preference · 2026-10-28 · rating

DeepSeek-V3

1,342.78 rating

Confidence interval: 1,332.26 – 1,353.29

3123 votes

Category
longer_query
Evaluated configuration
DeepSeek-V3
Evaluation type
human preference
Arena text preference / mathhuman preference · 2026-10-28 · rating

DeepSeek-V3

1,310.66 rating

Confidence interval: 1,299.90 – 1,321.41

2721 votes

Category
math
Evaluated configuration
DeepSeek-V3
Evaluation type
human preference
Arena text preference / multi_turnhuman preference · 2026-10-28 · rating

DeepSeek-V3

1,349.06 rating

Confidence interval: 1,339.23 – 1,358.88

3747 votes

Category
multi_turn
Evaluated configuration
DeepSeek-V3
Evaluation type
human preference
Arena text preference / non_englishhuman preference · 2026-10-28 · rating

DeepSeek-V3

1,315.43 rating

Confidence interval: 1,308.75 – 1,322.11

8904 votes

Category
non_english
Evaluated configuration
DeepSeek-V3
Evaluation type
human preference
Arena text preference / overallhuman preference · 2026-10-28 · rating

DeepSeek-V3

1,332.47 rating

Confidence interval: 1,327.74 – 1,337.20

21770 votes

Category
overall
Evaluated configuration
DeepSeek-V3
Evaluation type
human preference
Arena text preference / russianhuman preference · 2026-10-28 · rating

DeepSeek-V3

1,323.93 rating

Confidence interval: 1,312.51 – 1,335.35

2512 votes

Category
russian
Evaluated configuration
DeepSeek-V3
Evaluation type
human preference
Arena text preference / spanishhuman preference · 2026-10-28 · rating

DeepSeek-V3

1,352.44 rating

Confidence interval: 1,304.31 – 1,400.58

127 votes

Category
spanish
Evaluated configuration
DeepSeek-V3
Evaluation type
human preference
Arena text preference / chinesehuman preference · 2026-10-15 · rating

DeepSeek-V3

1,338.23 rating

Confidence interval: 1,322.88 – 1,353.58

1450 votes

Category
chinese
Evaluated configuration
deepseek-v3
Evaluation type
human preference
Arena text preference / codinghuman preference · 2026-10-15 · rating

DeepSeek-V3

1,324.81 rating

Confidence interval: 1,314.72 – 1,334.89

3280 votes

Category
coding
Evaluated configuration
deepseek-v3
Evaluation type
human preference
Arena text preference / creative_writinghuman preference · 2026-10-15 · rating

DeepSeek-V3

1,328.89 rating

Confidence interval: 1,318.86 – 1,338.91

3623 votes

Category
creative_writing
Evaluated configuration
deepseek-v3
Evaluation type
human preference
Arena text preference / englishhuman preference · 2026-10-15 · rating

DeepSeek-V3

1,345.71 rating

Confidence interval: 1,339.79 – 1,351.64

12866 votes

Category
english
Evaluated configuration
deepseek-v3
Evaluation type
human preference
Arena text preference / exclude_tieshuman preference · 2026-10-15 · rating

DeepSeek-V3

1,288.52 rating

Confidence interval: 1,281.22 – 1,295.82

14106 votes

Category
exclude_ties
Evaluated configuration
deepseek-v3
Evaluation type
human preference
Arena text preference / experthuman preference · 2026-10-15 · rating

DeepSeek-V3

1,305.45 rating

Confidence interval: 1,288.93 – 1,321.96

1236 votes

Category
expert
Evaluated configuration
deepseek-v3
Evaluation type
human preference
Arena text preference / frenchhuman preference · 2026-10-15 · rating

DeepSeek-V3

1,342.92 rating

Confidence interval: 1,305.39 – 1,380.44

242 votes

Category
french
Evaluated configuration
deepseek-v3
Evaluation type
human preference
Arena text preference / germanhuman preference · 2026-10-15 · rating

DeepSeek-V3

1,323.55 rating

Confidence interval: 1,298.37 – 1,348.72

493 votes

Category
german
Evaluated configuration
deepseek-v3
Evaluation type
human preference
Arena text preference / hard_promptshuman preference · 2026-10-15 · rating

DeepSeek-V3

1,311.87 rating

Confidence interval: 1,303.83 – 1,319.91

5408 votes

Category
hard_prompts
Evaluated configuration
deepseek-v3
Evaluation type
human preference
Arena text preference / hard_prompts_englishhuman preference · 2026-10-15 · rating

DeepSeek-V3

1,323.18 rating

Confidence interval: 1,313.58 – 1,332.79

3637 votes

Category
hard_prompts_english
Evaluated configuration
deepseek-v3
Evaluation type
human preference
Arena text preference / industry_business_and_management_and_financial_operationshuman preference · 2026-10-15 · rating

DeepSeek-V3

1,306.95 rating

Confidence interval: 1,294.59 – 1,319.31

2241 votes

Category
industry_business_and_management_and_financial_operations
Evaluated configuration
deepseek-v3
Evaluation type
human preference
Arena text preference / industry_entertainment_and_sports_and_mediahuman preference · 2026-10-15 · rating

DeepSeek-V3

1,323.91 rating

Confidence interval: 1,314.64 – 1,333.17

4191 votes

Category
industry_entertainment_and_sports_and_media
Evaluated configuration
deepseek-v3
Evaluation type
human preference
Arena text preference / industry_legal_and_governmenthuman preference · 2026-10-15 · rating

DeepSeek-V3

1,341.20 rating

Confidence interval: 1,324.73 – 1,357.68

1312 votes

Category
industry_legal_and_government
Evaluated configuration
deepseek-v3
Evaluation type
human preference
Arena text preference / industry_life_and_physical_and_social_sciencehuman preference · 2026-10-15 · rating

DeepSeek-V3

1,332.57 rating

Confidence interval: 1,322.91 – 1,342.24

3766 votes

Category
industry_life_and_physical_and_social_science
Evaluated configuration
deepseek-v3
Evaluation type
human preference
Arena text preference / industry_mathematicalhuman preference · 2026-10-15 · rating

DeepSeek-V3

1,313.82 rating

Confidence interval: 1,302.05 – 1,325.59

2326 votes

Category
industry_mathematical
Evaluated configuration
deepseek-v3
Evaluation type
human preference
Arena text preference / industry_medicine_and_healthcarehuman preference · 2026-10-15 · rating

DeepSeek-V3

1,345.24 rating

Confidence interval: 1,327.02 – 1,363.47

1060 votes

Category
industry_medicine_and_healthcare
Evaluated configuration
deepseek-v3
Evaluation type
human preference
Arena text preference / industry_software_and_it_serviceshuman preference · 2026-10-15 · rating

DeepSeek-V3

1,336.45 rating

Confidence interval: 1,328.05 – 1,344.85

5176 votes

Category
industry_software_and_it_services
Evaluated configuration
deepseek-v3
Evaluation type
human preference
Arena text preference / industry_writing_and_literature_and_languagehuman preference · 2026-10-15 · rating

DeepSeek-V3

1,331.17 rating

Confidence interval: 1,323.03 – 1,339.31

5913 votes

Category
industry_writing_and_literature_and_language
Evaluated configuration
deepseek-v3
Evaluation type
human preference
Arena text preference / instruction_followinghuman preference · 2026-10-15 · rating

DeepSeek-V3

1,314.90 rating

Confidence interval: 1,308.06 – 1,321.75

8606 votes

Category
instruction_following
Evaluated configuration
deepseek-v3
Evaluation type
human preference
Arena text preference / japanesehuman preference · 2026-10-15 · rating

DeepSeek-V3

1,266.28 rating

Confidence interval: 1,242.47 – 1,290.09

534 votes

Category
japanese
Evaluated configuration
deepseek-v3
Evaluation type
human preference
Arena text preference / koreanhuman preference · 2026-10-15 · rating

DeepSeek-V3

1,248.45 rating

Confidence interval: 1,213.26 – 1,283.63

295 votes

Category
korean
Evaluated configuration
deepseek-v3
Evaluation type
human preference
Arena text preference / longer_queryhuman preference · 2026-10-15 · rating

DeepSeek-V3

1,342.83 rating

Confidence interval: 1,332.31 – 1,353.35

3123 votes

Category
longer_query
Evaluated configuration
deepseek-v3
Evaluation type
human preference
Arena text preference / mathhuman preference · 2026-10-15 · rating

DeepSeek-V3

1,310.62 rating

Confidence interval: 1,299.87 – 1,321.37

2721 votes

Category
math
Evaluated configuration
deepseek-v3
Evaluation type
human preference
Arena text preference / multi_turnhuman preference · 2026-10-15 · rating

DeepSeek-V3

1,349.06 rating

Confidence interval: 1,339.23 – 1,358.88

3747 votes

Category
multi_turn
Evaluated configuration
deepseek-v3
Evaluation type
human preference
Arena text preference / non_englishhuman preference · 2026-10-15 · rating

DeepSeek-V3

1,315.53 rating

Confidence interval: 1,308.85 – 1,322.21

8904 votes

Category
non_english
Evaluated configuration
deepseek-v3
Evaluation type
human preference
Arena text preference / overallhuman preference · 2026-10-15 · rating

DeepSeek-V3

1,332.46 rating

Confidence interval: 1,327.73 – 1,337.19

21770 votes

Category
overall
Evaluated configuration
deepseek-v3
Evaluation type
human preference
Arena text preference / russianhuman preference · 2026-10-15 · rating

DeepSeek-V3

1,324.08 rating

Confidence interval: 1,312.65 – 1,335.50

2512 votes

Category
russian
Evaluated configuration
deepseek-v3
Evaluation type
human preference
Arena text preference / spanishhuman preference · 2026-10-15 · rating

DeepSeek-V3

1,352.37 rating

Confidence interval: 1,304.21 – 1,400.52

127 votes

Category
spanish
Evaluated configuration
deepseek-v3
Evaluation type
human preference

Run it locally

Inference · memory estimate

DeepSeek-V3

Not enough verified architecture information for a reliable VRAM estimate.

API cost

USD / month

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

DeepSeek-V3

DeepSeek-V3 / Pricing tiers 1
Provider / scopeRateCurrencyAmount / unit
DeepInfradeepseek-ai/DeepSeek-V3Standard InputUSD0.32 / 1M tokens
OutputUSD0.89 / 1M tokens

Capabilities

Context window
131,072 tokens
Maximum input
65,536 tokens*
Maximum output
8,192 tokens
Reasoning
No
Tool calling
Yes
Open weights
Yes
API mode
Chat*
Prompt caching
Yes*
Tool choice
Yes*
Model license
DeepSeek Model License

Input modalities

  • text

Output modalities

  • text

System configuration

Detailed specifications

Specifications · 14 fields
cache creation input token cost
0.0*
Scope and conditions
native_context
provider
deepseek*
mode
chat*
full_native_model_id
deepseek/deepseek-v3*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
cache read input token cost
7.0e-8*
Scope and conditions
native_context
provider
deepseek*
mode
chat*
full_native_model_id
deepseek/deepseek-v3*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
input cost per token
2.7e-7*
Scope and conditions
native_context
provider
deepseek*
mode
chat*
full_native_model_id
deepseek/deepseek-v3*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
input cost per token cache hit
7.0e-8*
Scope and conditions
native_context
provider
deepseek*
mode
chat*
full_native_model_id
deepseek/deepseek-v3*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
litellm provider
deepseek*
Scope and conditions
native_context
provider
deepseek*
mode
chat*
full_native_model_id
deepseek/deepseek-v3*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
max input tokens
65,536*
Scope and conditions
native_context
provider
deepseek*
mode
chat*
full_native_model_id
deepseek/deepseek-v3*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
max output tokens
8,192*
Scope and conditions
native_context
provider
deepseek*
mode
chat*
full_native_model_id
deepseek/deepseek-v3*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
max tokens
8,192*
Scope and conditions
native_context
provider
deepseek*
mode
chat*
full_native_model_id
deepseek/deepseek-v3*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
mode
chat*
Scope and conditions
native_context
provider
deepseek*
mode
chat*
full_native_model_id
deepseek/deepseek-v3*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
output cost per token
1.1e-6*
Scope and conditions
native_context
provider
deepseek*
mode
chat*
full_native_model_id
deepseek/deepseek-v3*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supports assistant prefill
Yes*
Scope and conditions
native_context
provider
deepseek*
mode
chat*
full_native_model_id
deepseek/deepseek-v3*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supports function calling
Yes*
Scope and conditions
native_context
provider
deepseek*
mode
chat*
full_native_model_id
deepseek/deepseek-v3*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supports prompt caching
Yes*
Scope and conditions
native_context
provider
deepseek*
mode
chat*
full_native_model_id
deepseek/deepseek-v3*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supports tool choice
Yes*
Scope and conditions
native_context
provider
deepseek*
mode
chat*
full_native_model_id
deepseek/deepseek-v3*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*

* Provisional data