BenchAtlas

Rankings / DeepSeek

Deepseek V4 Flash

released 2026-04-24

BenchAtlas Index

as of 2026-09-11
63.8
max effort
rank #121
7 families · 4 categories · medium
60.5
high effort
rank #150
7 families · 4 categories · medium
49.2
no reasoning
rank #245
7 families · 4 categories · medium

Benchmark evidence

65 results
agentic coding
Release 2026-06-25
tasks_counted=3 · livebench_version=2026-06-25
37.6 %independentLiveBench
τ²-Benchhigh effort95.6 %independentArtificial Analysis
τ²-Benchmax effort95.0 %independentArtificial Analysis
τ²-Benchno reasoning94.4 %independentArtificial Analysis
τ²-Bench
subset=banking
max effort30.9 %independentArtificial Analysis
τ²-Bench
subset=banking
high effort26.2 %independentArtificial Analysis
coding
Release 2026-06-25
tasks_counted=2 · livebench_version=2026-06-25
69.2 %independentLiveBench
SciCodemax effort45.3 %independentArtificial Analysis
SciCodehigh effort40.2 %independentArtificial Analysis
SciCodeno reasoning37.3 %independentArtificial Analysis
data analysis
Release 2026-06-25
tasks_counted=3 · livebench_version=2026-06-25
68.0 %independentLiveBench
external indices
AA Coding Indexmax effort56.2 pointsindependentArtificial Analysis
AA Coding Indexhigh effort52.0 pointsindependentArtificial Analysis
Intelligence Index v4.1high effort24.8 pointsindependentArtificial Analysis
Intelligence Index v4.1max effort24.6 pointsindependentArtificial Analysis
Intelligence Index v4.1no reasoning18.9 pointsindependentArtificial Analysis
ECIhigh effort146.3 pointsindependentEpoch AI Benchmarking Hub
ECImax effort146.3 pointsindependentEpoch AI Benchmarking Hub
ECI146.3 pointsindependentEpoch AI Benchmarking Hub
ECIno reasoning146.3 pointsindependentEpoch AI Benchmarking Hub
human preference
Coding (style control)older version
arena=text · category=coding · style_control=true
1483.0
1476.6–1489.4
communityLMArena Leaderboard Dataset
Chinese (style control)older version
arena=text · category=chinese · style_control=true
1472.7
1460.2–1485.1
communityLMArena Leaderboard Dataset
Industry Software And It Services (style control)older version
arena=text · category=industry_software_and_it_services · style_control=true
1471.0
1465.4–1476.7
communityLMArena Leaderboard Dataset
Hard Prompts English (style control)older version
arena=text · category=hard_prompts_english · style_control=true
1464.1
1457.9–1470.3
communityLMArena Leaderboard Dataset
Industry Medicine And Healthcare (style control)older version
arena=text · category=industry_medicine_and_healthcare · style_control=true
1459.1
1448.1–1470.0
communityLMArena Leaderboard Dataset
Hard Prompts (style control)older version
arena=text · category=hard_prompts · style_control=true
1459.1
1454.2–1463.9
communityLMArena Leaderboard Dataset
Expert (style control)older version
arena=text · category=expert · style_control=true
1458.2
1449.0–1467.3
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
1458.1
1450.5–1465.7
communityLMArena Leaderboard Dataset
Multi Turn (style control)older version
arena=text · category=multi_turn · style_control=true
1452.4
1444.9–1459.8
communityLMArena Leaderboard Dataset
Industry Legal And Government (style control)older version
arena=text · category=industry_legal_and_government · style_control=true
1449.1
1438.9–1459.2
communityLMArena Leaderboard Dataset
Longer Query (style control)older version
arena=text · category=longer_query · style_control=true
1448.2
1442.5–1454.0
communityLMArena Leaderboard Dataset
English (style control)older version
arena=text · category=english · style_control=true
1445.2
1439.8–1450.6
communityLMArena Leaderboard Dataset
French (style control)older version
arena=text · category=french · style_control=true
1444.0
1428.5–1459.5
communityLMArena Leaderboard Dataset
Polish (style control)older version
arena=text · category=polish · style_control=true
1439.6
1420.6–1458.6
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
1437.7
1430.6–1444.8
communityLMArena Leaderboard Dataset
Overall (style control)
arena=text · category=overall · style_control=true
1435.8
1431.6–1439.9
communityLMArena Leaderboard Dataset
Russian (style control)older version
arena=text · category=russian · style_control=true
1433.2
1424.4–1442.1
communityLMArena Leaderboard Dataset
Spanish (style control)older version
arena=text · category=spanish · style_control=true
1433.0
1416.6–1449.4
communityLMArena Leaderboard Dataset
Industry Mathematical (style control)older version
arena=text · category=industry_mathematical · style_control=true
1432.1
1420.1–1444.0
communityLMArena Leaderboard Dataset
Exclude Ties (style control)older version
arena=text · category=exclude_ties · style_control=true
1431.2
1425.8–1436.6
communityLMArena Leaderboard Dataset
Instruction Following (style control)older version
arena=text · category=instruction_following · style_control=true
1428.2
1422.2–1434.1
communityLMArena Leaderboard Dataset
Math (style control)older version
arena=text · category=math · style_control=true
1425.6
1413.4–1437.8
communityLMArena Leaderboard Dataset
German (style control)older version
arena=text · category=german · style_control=true
1423.9
1402.4–1445.4
communityLMArena Leaderboard Dataset
Non English (style control)older version
arena=text · category=non_english · style_control=true
1423.1
1418.0–1428.1
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
1418.0
1411.2–1424.7
communityLMArena Leaderboard Dataset
Creative Writing (style control)older version
arena=text · category=creative_writing · style_control=true
1408.1
1400.3–1415.8
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
1405.3
1398.1–1412.4
communityLMArena Leaderboard Dataset
Japanese (style control)older version
arena=text · category=japanese · style_control=true
1397.1
1371.8–1422.4
communityLMArena Leaderboard Dataset
Korean (style control)older version
arena=text · category=korean · style_control=true
1392.9
1371.5–1414.3
communityLMArena Leaderboard Dataset
knowledge science
GPQA Diamond
implementation=artificial-analysis
max effort89.4 %independentArtificial Analysis
GPQA Diamond
implementation=artificial-analysis
high effort86.7 %independentArtificial Analysis
GPQA Diamond
implementation=artificial-analysis
no reasoning71.6 %independentArtificial Analysis
Humanity's Last Exam
implementation=artificial-analysis
max effort34.8 %independentArtificial Analysis
Humanity's Last Exam
implementation=artificial-analysis
high effort30.3 %independentArtificial Analysis
Humanity's Last Exam
implementation=artificial-analysis
no reasoning7.8 %independentArtificial Analysis
language
Release 2026-06-25
tasks_counted=3 · livebench_version=2026-06-25
70.1 %independentLiveBench
long context instruction
AA-LCRmax effort74.3 %independentArtificial Analysis
AA-LCRhigh effort72.0 %independentArtificial Analysis
AA-LCRno reasoning41.7 %independentArtificial Analysis
IFBenchmax effort79.2 %independentArtificial Analysis
IFBenchhigh effort73.5 %independentArtificial Analysis
IFBenchno reasoning47.2 %independentArtificial Analysis
Release 2026-06-25
tasks_counted=4 · livebench_version=2026-06-25
63.1 %independentLiveBench
reasoning math
Release 2026-06-25
tasks_counted=4 · livebench_version=2026-06-25
79.7 %independentLiveBench
Release 2026-06-25
tasks_counted=4 · livebench_version=2026-06-25
70.6 %independentLiveBench

Agent + model results

systems, not bare-model scores
agent + model Artificial Analysis harness + Deepseek V4 FlashTerminal-Bench 2.161.8 %independentArtificial Analysis
agent + model Artificial Analysis harness + Deepseek V4 FlashTerminal-Bench 2.156.9 %independentArtificial Analysis
agent + model Artificial Analysis harness + Deepseek V4 FlashTerminal-Bench Hard38.6 %independentArtificial Analysis
agent + model Artificial Analysis harness + Deepseek V4 FlashTerminal-Bench Hard35.6 %independentArtificial Analysis
agent + model Artificial Analysis harness + Deepseek V4 FlashTerminal-Bench Hard34.1 %independentArtificial Analysis

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