Atlas

Compare Models

DeepSeek V4.1 Flash vs Inkling 2 Small

Inkling 2 Small leads by 0.7 capability points (≈95% interval −1.2 to +2.7): statistically tied · P(higher) ≈ 77%

Wins 46 of 70 shared benchmarks (23 lost, 1 tied)

Capability vs. relative cost
DeepSeek V4.1 FlashInkling 2 SmallClosed 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 FlashvsInkling 2 Smallcapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins23 – 46 · 70 shared, 1 tied

Comparing matching setups only (42 benchmarks excluded) · suites counted once
← DeepSeek V4.1 FlashInkling 2 Small →

DeepSeek V4.1 FlashvsInkling 2 Smallwins by categoryData Nov 15, 2026Data: Vals AI, Artificial Analysis, MathArena and 7 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Inkling 2 Small minus DeepSeek V4.1 Flash.

EnterpriseBench: CoreCraftProfessional Work41.571.8+30.3
ChartographyMultimodal11.636.7+25.1
SWE-bench MultilingualCode98.281.2−17.0
AA-AnalystAgent (pass^5)Professional Work35.051.3+16.2
Terminal-Bench 4.0 (Artificial Analysis)Code26.813.1−13.6
GDP.pdf (Artificial Analysis, All-pass)Professional Work12.825.2+12.4
APEX-Accounting 1.1Finance & Accounting50.162.0+11.9
APEX-Agents 1.1Professional Work50.560.9+10.4
Humanity's Last Exam DiamondGeneral QA19.128.7+9.6
ArXivMath 09/2026 (Native Harnesses)Math40.631.1−9.4
Terminal-Bench 2.1Code90.681.3−9.3
Finance Agent v2.1 - ComparablesFinance & Accounting55.865.0+9.2
Finance Agent v2.1 - PrecedentsFinance & Accounting37.947.1+9.2
Legal Research Bench FamilyLegal13.622.7+9.1
Legal Research Bench ImmigrationLegal36.427.3−9.1
ARC-AGI-2General QA72.981.7+8.8
BrokenArXiv 09/2026 (Native Harnesses)Math54.763.2+8.5
BrokenArXiv 08/2026 (Native Harnesses)Math54.863.1+8.3
Finance Agent v2.1 - Market AnalysisFinance & Accounting67.159.0−8.1
SAGE v1.1 - Linear AlgebraMath72.780.5+7.8
Legal Research Bench Administrative/RegulatoryLegal53.060.6+7.6
Terminal-Bench 4.0Code19.712.6−7.1
ArXivMath 08/2026 (Native Harnesses)Math43.936.8−7.0
Legal Research Bench Business/CommercialLegal39.546.5+7.0
AA-Briefcase v1.1 EloAgents14151527+6.1
Arena Code WebDev — HTMLCode16181660+5.8
Excel Modeling Benchmark v1.1 MiscFinance & Accounting62.568.3+5.7
Artificial Analysis Openness IndexIndexes44.438.9−5.6
ComplexConstraintsChat & Writing49.744.5−5.2
Excel Modeling Benchmark v1.1 CompsFinance & Accounting68.073.1+5.1
Finance Agent v2.1 - General Qualitative AnalysisFinance & Accounting73.078.0+5.0
Finance Agent v2.1 - All PassFinance & Accounting51.756.6+4.9
Finance Agent v2.1 - Financial ModelingFinance & Accounting46.842.4−4.4
MysteryMechanismScience21.225.2+4.1
Finance Agent v2.1 - Disclosure AnalysisFinance & Accounting76.980.8+3.9
Arena Code WebDev — ReactCode16301659+3.8
SAGE v1.1 - Partial Differential EquationsMath90.394.0+3.7
Legal Research Bench Civil LitigationLegal48.151.9+3.7
Legal Research Bench CriminalLegal33.336.8+3.5
Excel Modeling Benchmark v1.1 DataroomFinance & Accounting79.983.4+3.5
Legal Research Bench Weighted PassLegal84.587.9+3.4
Terminal-Bench-Science 0.1Science9.05.7−3.3
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine23.920.6−3.3
Legal Research Bench HealthLegal73.370.0−3.3
SWE-bench Multimodal (Vals v1, 480 Tasks)Code46.543.1−3.3
Tax Agent BenchFinance & Accounting62.565.8+3.3
Vibe Code Bench v1.1Code84.781.5−3.2
Vals Science IndexIndexes29.632.7+3.0
Terminal-Bench Science 0.1 (Vals)Science0.02.9+2.9
Finance Agent v2.1 - General Quantitative AnalysisFinance & Accounting76.579.2+2.7
Legal Research Bench Constitutional/Civil RightsLegal36.839.5+2.6
Excel Modeling Benchmark v1.1 DCFFinance & Accounting59.962.2+2.4
BioMysteryBench (Vals Terminus 2)Science67.870.0+2.2
Apex (MathArena)Math65.663.5−2.1
Excel Modeling Benchmark v1.1 M&AFinance & Accounting75.773.7−2.1
ARC-AGI-1 Semi-PrivateGeneral QA94.596.5+2.0
Excel Modeling Benchmark v1.1 OperatingFinance & Accounting71.970.0−1.9
Excel Modeling Benchmark v1.1 ScratchFinance & Accounting65.267.1+1.8
Excel Modeling Benchmark v1.1 TemplateFinance & Accounting70.271.6+1.4
GDPval-AA v2.1Professional Work16001630+1.3
SRE Bench — Fully SolvedCode0.81.9+1.1
CyberBench v1.1 OverallCybersecurity73.772.7−1.0
Finance Agent v2.1 - AdjustmentsFinance & Accounting66.667.4+0.8
SWE-bench VerifiedCode82.982.2−0.7
Finance Agent v2.1 - Earnings AnalysisFinance & Accounting88.388.9+0.7
Excel Modeling Benchmark v1.1 LBOFinance & Accounting56.455.9−0.5
DeepSWE v1.1Code72.672.1−0.5
SciCode 1.0.1 (Artificial Analysis)Code51.952.3+0.5
SAGE v1.1 - Computer ScienceMath91.191.2+0.2
ProgramBenchCode0.50.50.0

DeepSeek V4.1 FlashvsInkling 2 Smallnormalised score marginsData Nov 15, 2026Data: Vals AI, Artificial Analysis, MathArena and 7 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…