Atlas

Compare Models

DeepSeek V4.1 Flash vs DeepSeek V4.1 Flash 1017

DeepSeek V4.1 Flash 1017 leads by 1.6 capability points (≈95% interval −0.7 to +3.8): statistically tied · P(higher) ≈ 91%

Wins 86 of 115 shared benchmarks (26 lost, 3 tied)

Capability vs. relative cost
DeepSeek V4.1 FlashDeepSeek V4.1 Flash 1017Closed weightsOpen weights

Each line runs through one model’s priced reasoning settings, cheapest to most expensive; grey points are other models’ priced settings in the same range. Results whose reasoning effort the source didn’t report are left out wherever a model has named effort levels. Capability and relative cost are the published estimates (relative cost is cost per task as a multiple of GPT-5 (high)); hover a setting, or use the arrow keys, to set it beside the other model’s same setting.

DeepSeek V4.1 FlashvsDeepSeek V4.1 Flash 1017capability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins26 – 86 · 115 shared, 3 tied

Comparing matching setups only (1 benchmark excluded) · suites counted once
← DeepSeek V4.1 FlashDeepSeek V4.1 Flash 1017 →

DeepSeek V4.1 FlashvsDeepSeek V4.1 Flash 1017wins by categoryData Nov 15, 2026Data: Artificial Analysis, Arena, Vals AI and 26 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is DeepSeek V4.1 Flash 1017 minus DeepSeek V4.1 Flash.

CyberGym-E2E-AA (Artificial Analysis)Cybersecurity22.976.3+53.4
APEX-Agents 1.1Professional Work50.579.4+28.9
AA-PersonalityBench Transparency indexChat & Writing60.086.7+26.9
FrontierMath Tier 4 (v2) - PrivateMath26.846.3+19.5
EnterpriseBench: CoreCraftProfessional Work41.558.5+17.0
AA-Briefcase v1.1 EloAgents11181425+16.8
Bug Hunt BenchCode20.637.3+16.6
Concept Synth INDUCTION Challenge64 Round 1General QA17.232.8+15.6
AA-PersonalityBench Warmth indexChat & Writing41.027.7−13.5
VoxelBench Text RatingGames18612047+13.2
Tax Agent BenchFinance & Accounting62.574.9+12.4
AA-PersonalityBench Organization indexChat & Writing10088.6−11.5
τ³-Banking (Fireworks, banking_knowledge)Agents35.746.8+11.1
Wiz Cyber Model Arena v3.9Cybersecurity50.260.3+10.1
Concept Synth INDUCTION Challenge100 Round 1General QA11.021.0+10.0
CritPt (avg@5)Science14.324.0+9.7
AA-PersonalityBench Expressiveness indexChat & Writing61.952.4−9.6
APEX-Accounting 1.1Finance & Accounting50.159.3+9.2
ProgramBench - almost solved rollout percentageCode20.329.1+8.8
Code Migration v1.1Code58.967.5+8.6
LMArena Agent Arena Code Net Improvement (Steering Burden)Agents2.76.5+8.5
BrokenArXiv 08/2026 (Native Harnesses)Math54.846.7−8.0
AA-PersonalityBench Informativeness indexChat & Writing79.171.4−7.7
LMArena Agent Arena Work Net Improvement (Steering Burden)Agents2.55.6+7.4
Terminal-Bench 4.0 (Artificial Analysis)Code26.833.8+7.1
ALE-BenchCode18292012+6.5
DeepsecBench-AA (Artificial Analysis)Cybersecurity31.224.9−6.3
Mystery Game PuzzlesGames43.049.0+6.0
GBENCH One-shot Coding GScoreGames48.053.8+5.8
Humanity's Last ExamGeneral QA36.842.6+5.8
AA-PersonalityBench Imagination indexChat & Writing31.537.2+5.8
AutomationBench 1.0.6 (provider-reported)Agents54.860.5+5.7
Artificial Analysis Intelligence Index v4.3.2Indexes39.545.1+5.6
LMArena Code Arena WebDevCode16211664+5.6
TCSExamMath40.446.0+5.6
Aikido Known-CVE Detection (32 CVEs) single-run averageCybersecurity65.670.8+5.2
ARC-AGI-2General QA72.978.1+5.1
BrokenArXiv 09/2026 (Native Harnesses)Math54.749.7−5.0
AA-Omniscience IndexIndexes-5.34.7+5.0
ArXivMath 09/2026 (Native Harnesses)Math40.645.3+4.7
Excel Modeling BenchmarkFinance & Accounting57.261.8+4.6
Design Arena Overall Code CategoriesCode13161289−4.5
GDP.pdf (Artificial Analysis, All-pass)Professional Work12.817.2+4.4
Chartography (with tools)Multimodal78.983.0+4.1
SciCode 1.0.1 (Artificial Analysis)Code51.955.9+4.1
FoodTruck BenchGames5017953289+4.0
HealthBench Professional (Length-Adjusted)Health & Medicine47.851.7+4.0
FrontierMath Tiers 1-3 (v2) - PrivateMath67.471.2+3.9
Legal Research BenchLegal41.345.2+3.8
Terminal-Bench-Science 0.1Science9.012.9+3.8
Vals Finance IndexIndexes57.160.9+3.8
τ³-Banking (Mercor run, bm25_grep retrieval)Agents37.841.6+3.8
Agents' Last Exam - release-reported pass rate, subset unspecifiedAgents31.835.4+3.6
MMMU ProMultimodal65.669.2+3.6
Finance Agent v2Finance & Accounting53.557.0+3.5
WeirdML v3Science5.89.2+3.4
ARC-AGI-3 (Semi-Private)General QA2.75.8+3.1
Vibe Code Bench v1.1Code84.787.7+3.0
LMArena Agent Arena Code Steering BurdenAgents2.38.2+3.0
BabyVision with toolsMultimodal89.692.5+2.9
LMArena Agent Arena Code Praise vs ComplaintAgents3.08.8+2.9
LMArena Agent Arena Work Praise vs ComplaintAgents3.79.3+2.8
Hebbia Financial Services BenchmarkFinance & Accounting63.160.3−2.8
APEX-1 (Mercor, mean score)Professional Work65.568.3+2.8
APEX-SWECode51.949.1−2.8
DeepSWE v1.1Code74.277.0+2.8
NL2Repo - release-reported repository generation scoreCode65.468.2+2.8
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine23.926.7+2.8
Bedside BenchHealth & Medicine82.985.7+2.8
LMArena Agent Arena Work Bash ReliabilityAgents0.15.4+2.6
LMArena Agent Arena Chat Bash ReliabilityAgents-1.04.2+2.6
BuseyBench SVGMultimodal4.14.3+2.6
LMArena Agent Arena Code Bash ReliabilityAgents-0.24.9+2.5
Harvey's Legal Agent BenchmarkLegal6.79.2+2.5
Finance Agent v2.1Finance & Accounting65.462.9−2.5
Epoch Capabilities Index (ECI)Indexes156159+2.3
DAXBench v2Code75.577.7+2.1
LMArena Agent Arena Chat Net Improvement (Steering Burden)Agents1.62.5+2.0
AA-LCR v1.1Search55.357.3+2.0
Chess Puzzles (task version 2.0)Games48.046.0−2.0
DualEntry Accounting AI BenchmarkFinance & Accounting84.282.2−2.0
ZeroBench (Main questions, pass@5, with tools)Multimodal49.051.0+2.0
AA-PersonalityBench Conviction indexChat & Writing96.294.3−1.9
AA-PersonalityBench Slop indexChat & Writing37.239.1+1.9
LMArena Agent Arena Work Steering BurdenAgents1.55.2+1.8
Humanity's Last Exam DiamondGeneral QA19.120.8+1.7
GameBench 2Games68.469.9+1.5
LMArena Agent Arena Code Confirmed SuccessAgents6.69.4+1.4
Excel Modeling Benchmark v1.1Finance & Accounting67.869.2+1.4
Code MigrationCode45.644.2−1.4
Legacy-Bench (Fireworks, avg@3)Code57.756.3−1.3
Artificial Analysis Progress Index v1Indexes320324+1.2
LMArena Agent Arena Chat Praise vs ComplaintAgents3.10.7−1.2
STEMBench120 v1.0.0 (closed-book)Science2.52.5+1.2
LiveBench 2.0Agents78.079.2+1.2
OTIS Mock AIME 2024-2025Math98.399.4+1.1
LMArena Agent Arena Chat Confirmed SuccessAgents4.97.1+1.1
LMArena Agent Arena Work Confirmed SuccessAgents5.77.5+0.9
CWE-Bench-AA (Artificial Analysis)Cybersecurity68.369.2+0.8
LMArena Agent Arena Chat Steering BurdenAgents-0.50.8+0.6
CyberGym (pass@3)Cybersecurity94.895.4+0.5
SimpleQA Verified (Epoch anti-abstention)General QA42.942.5−0.4
AutomationBench-AA 1.0.6Agents68.968.5−0.4
FrontierCode 1.1 Main (Artificial Analysis)Code44.043.7−0.4
Political Science LLM BenchmarkProfessional Work66.566.2−0.3
Arena Image-to-WebDevCode16101609−0.3
Access to Justice BenchmarkLegal72.973.1+0.2
Terminal-Bench 4.0Code31.231.0−0.2
LMArena Agent Arena Code Tool HallucinationAgents0.50.3−0.1
Epiq LAER CorporateBenchProfessional Work75.375.2−0.1
LMArena Agent Arena Work Tool HallucinationAgents0.50.4−0.1
LMArena Agent Arena Chat Tool HallucinationAgents0.30.30.0
ARC-AGI-1 Semi-PrivateGeneral QA94.594.50.0
Artificial Analysis Openness IndexIndexes44.444.40.0
ArXivMath 08/2026 (Native Harnesses)Math43.943.90.0

DeepSeek V4.1 FlashvsDeepSeek V4.1 Flash 1017normalised score marginsData Nov 15, 2026Data: Artificial Analysis, Arena, Vals AI and 26 othersVillage LLM Atlasatlas.village.dev

Win counts use each benchmark’s declared score direction. Represented suite slices and the derived Capability score are excluded from headline totals. Missing or incompatible setups are excluded, never counted as losses.
Loading Atlas data…