BenchAtlas

Rankings / DeepSeek

Deepseek V3 2

released 2025-12-01

BenchAtlas Index

as of 2026-09-11
58.7
thinking
rank #172
10 families · 5 categories · high
49.8
no reasoning
rank #241
10 families · 5 categories · high

Benchmark evidence

60 results
agentic coding
τ²-Benchthinking90.6 %independentArtificial Analysis
τ²-Benchno reasoning79.0 %independentArtificial Analysis
τ²-Bench
subset=banking
thinking18.8 %independentArtificial Analysis
coding
LiveCodeBench v6
implementation=artificial-analysis
thinking86.2 %independentArtificial Analysis
LiveCodeBench v6
implementation=artificial-analysis
no reasoning59.3 %independentArtificial Analysis
SciCodethinking38.9 %independentArtificial Analysis
SciCodeno reasoning38.7 %independentArtificial Analysis
external indices
AA Coding Indexthinking44.2 pointsindependentArtificial Analysis
Intelligence Index v4.1thinking21.5 pointsindependentArtificial Analysis
Intelligence Index v4.1no reasoning16.0 pointsindependentArtificial Analysis
AA Math Indexthinking92.0 pointsindependentArtificial Analysis
AA Math Indexno reasoning59.0 pointsindependentArtificial Analysis
human preference
Coding (style control)older version
arena=text · category=coding · style_control=true
1469.5
1463.1–1475.9
communityLMArena Leaderboard Dataset
Hard Prompts English (style control)older version
arena=text · category=hard_prompts_english · style_control=true
1458.1
1451.7–1464.4
communityLMArena Leaderboard Dataset
Chinese (style control)older version
arena=text · category=chinese · style_control=true
1458.0
1447.4–1468.6
communityLMArena Leaderboard Dataset
Industry Software And It Services (style control)older version
arena=text · category=industry_software_and_it_services · style_control=true
1457.5
1452.2–1462.9
communityLMArena Leaderboard Dataset
Industry Life And Physical And Social Science (style control)older version
arena=text · category=industry_life_and_physical_and_social_science · style_control=true
1451.4
1444.1–1458.8
communityLMArena Leaderboard Dataset
Expert (style control)older version
arena=text · category=expert · style_control=true
1448.0
1437.2–1458.6
communityLMArena Leaderboard Dataset
Hard Prompts (style control)older version
arena=text · category=hard_prompts · style_control=true
1447.2
1442.6–1451.7
communityLMArena Leaderboard Dataset
Industry Medicine And Healthcare (style control)older version
arena=text · category=industry_medicine_and_healthcare · style_control=true
1442.6
1431.3–1453.9
communityLMArena Leaderboard Dataset
French (style control)older version
arena=text · category=french · style_control=true
1442.4
1424.1–1460.6
communityLMArena Leaderboard Dataset
Longer Query (style control)older version
arena=text · category=longer_query · style_control=true
1441.2
1435.4–1447.1
communityLMArena Leaderboard Dataset
English (style control)older version
arena=text · category=english · style_control=true
1437.7
1432.6–1442.8
communityLMArena Leaderboard Dataset
Industry Mathematical (style control)older version
arena=text · category=industry_mathematical · style_control=true
1436.0
1423.2–1448.9
communityLMArena Leaderboard Dataset
Industry Legal And Government (style control)older version
arena=text · category=industry_legal_and_government · style_control=true
1432.7
1422.2–1443.2
communityLMArena Leaderboard Dataset
Math (style control)older version
arena=text · category=math · style_control=true
1428.6
1417.6–1439.6
communityLMArena Leaderboard Dataset
Multi Turn (style control)older version
arena=text · category=multi_turn · style_control=true
1428.1
1420.9–1435.3
communityLMArena Leaderboard Dataset
Overall (style control)
arena=text · category=overall · style_control=true
1425.3
1421.7–1428.8
communityLMArena Leaderboard Dataset
Russian (style control)older version
arena=text · category=russian · style_control=true
1421.9
1413.9–1429.9
communityLMArena Leaderboard Dataset
Industry Business And Management And Financial Operations (style control)older version
arena=text · category=industry_business_and_management_and_financial_operations · style_control=true
1421.6
1414.7–1428.5
communityLMArena Leaderboard Dataset
Instruction Following (style control)older version
arena=text · category=instruction_following · style_control=true
1420.4
1414.6–1426.2
communityLMArena Leaderboard Dataset
Spanish (style control)older version
arena=text · category=spanish · style_control=true
1419.9
1403.7–1436.0
communityLMArena Leaderboard Dataset
Exclude Ties (style control)older version
arena=text · category=exclude_ties · style_control=true
1416.6
1411.6–1421.6
communityLMArena Leaderboard Dataset
German (style control)older version
arena=text · category=german · style_control=true
1413.8
1393.1–1434.5
communityLMArena Leaderboard Dataset
Industry Writing And Literature And Language (style control)older version
arena=text · category=industry_writing_and_literature_and_language · style_control=true
1411.0
1404.6–1417.4
communityLMArena Leaderboard Dataset
Non English (style control)older version
arena=text · category=non_english · style_control=true
1408.9
1404.5–1413.3
communityLMArena Leaderboard Dataset
Polish (style control)older version
arena=text · category=polish · style_control=true
1402.7
1386.1–1419.3
communityLMArena Leaderboard Dataset
Creative Writing (style control)older version
arena=text · category=creative_writing · style_control=true
1401.3
1393.5–1409.1
communityLMArena Leaderboard Dataset
Industry Entertainment And Sports And Media (style control)older version
arena=text · category=industry_entertainment_and_sports_and_media · style_control=true
1396.1
1389.1–1403.1
communityLMArena Leaderboard Dataset
Korean (style control)older version
arena=text · category=korean · style_control=true
1368.8
1347.2–1390.4
communityLMArena Leaderboard Dataset
knowledge science
GPQA Diamond
implementation=artificial-analysis
thinking84.0 %independentArtificial Analysis
GPQA Diamond
implementation=artificial-analysis
no reasoning75.1 %independentArtificial Analysis
Humanity's Last Exam
implementation=artificial-analysis
thinking24.6 %independentArtificial Analysis
Humanity's Last Exam
implementation=artificial-analysis
no reasoning11.2 %independentArtificial Analysis
MMLU-Pro
implementation=artificial-analysis
thinking86.2 %independentArtificial Analysis
MMLU-Pro
implementation=artificial-analysis
no reasoning83.7 %independentArtificial Analysis
long context instruction
AA-LCRthinking73.3 %independentArtificial Analysis
AA-LCRno reasoning45.7 %independentArtificial Analysis
IFBenchthinking60.7 %independentArtificial Analysis
IFBenchno reasoning49.0 %independentArtificial Analysis
reasoning math
AIME
year=2025 · implementation=artificial-analysis
thinking92.0 %independentArtificial Analysis
AIME
year=2025 · implementation=artificial-analysis
no reasoning59.0 %independentArtificial Analysis
ARC-AGI-1older version
split=public_eval · model_type=Base LLM
61.6 %independentARC Prize Leaderboard
ARC-AGI-1older version
split=semi_private · model_type=Base LLM
57.0 %independentARC Prize Leaderboard
ARC-AGI-2
split=semi_private · model_type=Base LLM
4.0 %independentARC Prize Leaderboard
ARC-AGI-2
split=public_eval · model_type=Base LLM
3.9 %independentARC Prize Leaderboard
ARC-AGI-2
split=public_eval · model_type=Base LLM
0.1 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-2
split=semi_private · model_type=Base LLM
0.1 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-1older version
split=semi_private · model_type=Base LLM
0.1 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-1older version
split=public_eval · model_type=Base LLM
0.1 usd_per_taskindependentARC Prize Leaderboard

Agent + model results

systems, not bare-model scores
agent + model mini-SWE-agent + Deepseek V3 2SWE-bench Verified70.0 %unverifiedSWE-bench Leaderboard
agent + model mini-SWE-agent + Deepseek V3 2SWE-bench bash-only70.0 %unverifiedSWE-bench Leaderboard
agent + model mini-SWE-agent + Deepseek V3 2SWE-bench Multilingual59.0 %communitySWE-bench Leaderboard
agent + model mini-SWE-agent + Deepseek V3 2SWE-bench Verified0.4 usd_per_taskunverifiedSWE-bench Leaderboard
agent + model mini-SWE-agent + Deepseek V3 2SWE-bench bash-only0.4 usd_per_taskunverifiedSWE-bench Leaderboard
agent + model mini-SWE-agent + Deepseek V3 2SWE-bench Multilingual0.4 usd_per_taskcommunitySWE-bench Leaderboard
agent + model Artificial Analysis harness + Deepseek V3 2Terminal-Bench 2.146.8 %independentArtificial Analysis
agent + model terminus-2 + Deepseek V3 2Terminal-Bench 2.039.5 %unverifiedTerminal-Bench Leaderboard
agent + model Artificial Analysis harness + Deepseek V3 2Terminal-Bench Hard35.6 %independentArtificial Analysis
agent + model Artificial Analysis harness + Deepseek V3 2Terminal-Bench Hard32.6 %independentArtificial Analysis

These scores measure the whole agent system (scaffold, tools, budgets) — they are never merged into the bare model’s numbers.

Related variants

Deepseek V3 0324Deepseek V3Deepseek V3 1