Atlas

Compare Models

Claude Sonnet 5.5 vs DeepSeek V4.1 Flash 1017

Claude Sonnet 5.5 leads by 5.7 capability points (≈95% interval +3.1 to +8.3): a clear lead over DeepSeek V4.1 Flash 1017 · P(higher) ≈ 100%

Wins 112 of 126 shared benchmarks (14 lost)

Capability vs. relative cost
Claude Sonnet 5.5DeepSeek 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.

Claude Sonnet 5.5vsDeepSeek V4.1 Flash 1017capability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins112 – 14 · 126 shared

Comparing matching setups only (6 benchmarks excluded) · suites counted once
← Claude Sonnet 5.5DeepSeek V4.1 Flash 1017 →

Claude Sonnet 5.5vsDeepSeek V4.1 Flash 1017wins by categoryData Nov 15, 2026Data: Artificial Analysis, SecureBio, Surge AI and 36 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 Claude Sonnet 5.5.

HardGeoBench Avg DistanceMultimodal6833201−77.7
ARC-AGI-3 (Semi-Private)General QA72.45.8−66.6
LisanBench Difficulty-WeightedGames29837733+50.9
AA-PersonalityBench Imagination indexChat & Writing84.837.2−48.1
AA-PersonalityBench Warmth indexChat & Writing73.327.7−46.2
HardGeoBench Country AccuracyMultimodal58.815.0−43.7
HardGeoBench <=100miMultimodal50.07.5−42.5
Terminal-Bench-Science 0.1Science53.312.9−40.5
RiemannBench (November 2026 rescoring)Math76.838.4−38.4
AA-PersonalityBench Conviction indexChat & Writing58.194.3+36.5
HoneyBench v0.2 (hack rate)Safety41.676.4−34.8
FrontierMath Tier 4 (v2) - PrivateMath80.546.3−34.1
HardGeoBench <=10miMultimodal36.33.8−32.5
Humanity's Last Exam DiamondGeneral QA53.320.8−32.5
DAYJOB: Data ScienceProfessional Work35.33.3−32.0
Terminal-Bench 4.0Code61.831.0−30.8
Terminal-Bench 4.0 (Artificial Analysis)Code63.633.8−29.8
Arena Image-to-WebDevCode17441609−29.4
DAYJOB: Finance (October 2026 rescoring)Finance & Accounting40.512.5−28.0
DAYJOB: Healthcare (October 2026 rescoring)Health & Medicine34.06.0−28.0
DAYJOB: AccountingFinance & Accounting35.27.6−27.6
FrontierML ($200 wallet)Agents50.423.5−27.0
LMArena Agent Arena Net Improvement (Steering Burden)Agents16.02.0−26.3
Artificial Analysis Intelligence Index v5.0Indexes54.228.5−25.7
Code MigrationCode69.844.2−25.6
CritPt v2 (agentic, avg@5)Science75.750.6−25.1
HANDBOOK.md Agents v1.1Professional Work47.724.0−23.7
FrontierCode 2.0 - Pass RateCode46.423.0−23.4
Concept Synth INDUCTION Challenge100 Round 1General QA44.021.0−23.0
LisanBench Path LengthGames1378721107+22.7
Android Bench 2.0Code40.017.3−22.7
FrontierCode 2.0 - ScoreCode42.220.2−22.0
Artificial Analysis Intelligence Index v5.0.1Indexes54.532.6−22.0
BuseyBench SVGMultimodal6.54.3−21.8
GameBench 2Games91.469.9−21.5
AA-PersonalityBench Expressiveness indexChat & Writing73.352.4−21.2
Artificial Analysis Engineering IndexIndexes58.537.6−20.9
DAYJOB: LawLegal27.06.2−20.8
Code Migration v1.1Code86.967.5−19.4
VIALS-AA (pass^3)Multimodal59.640.4−19.3
Bio Capabilities Index (SecureBio)Indexes175149−18.8
FrontierMath Tiers 1-3 (v2) - PrivateMath88.871.2−17.5
APEX-SWECode66.449.1−17.3
Concept Synth INDUCTION Challenge64 Round 1General QA50.032.8−17.2
WCB (SecureBio)Science66.549.6−16.9
HPCT (SecureBio)Science77.260.5−16.7
Mystery Game PuzzlesGames65.049.0−16.0
Artificial Analysis Economics IndexIndexes54.138.5−15.6
ALE-BenchCode24482012−15.5
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine41.726.7−15.0
ATLAS (Exa, 200-task subset)Search46.831.9−14.9
LMArena Agent Arena Praise vs ComplaintAgents30.51.2−14.7
Artificial Analysis Legal IndexIndexes52.237.7−14.4
APEX-1 (Mercor, mean score)Professional Work53.968.3+14.4
Humanity's Last ExamGeneral QA56.942.6−14.3
LMArena Code Arena WebDevCode17731664−14.3
Excel Modeling BenchmarkFinance & Accounting75.761.8−13.9
AA-Omniscience IndexIndexes32.34.7−13.8
VCT v2 (SecureBio)Science52.238.5−13.7
Artificial Analysis Finance & Accounting IndexIndexes55.842.2−13.6
GBENCH One-shot Coding GScoreGames67.453.8−13.6
AA-PersonalityBench Transparency indexChat & Writing10086.7−13.5
AA-PersonalityBench Informativeness indexChat & Writing84.871.4−13.5
MKA (Artificial Analysis)Professional Work52.038.5−13.4
Chess Puzzles (task version 2.0)Games59.046.0−13.0
Artificial Analysis Healthcare & Medical IndexIndexes54.241.3−12.9
τ³-Banking (Mercor run, bm25_grep retrieval)Agents54.041.6−12.4
EnterpriseBench: CoreCraftProfessional Work70.858.5−12.3
VIALS-AA (average of 3)Multimodal65.453.2−12.2
AA-Briefcase v1.1 EloAgents16341416−11.9
Bug Hunt BenchCode48.937.3−11.6
AA-PersonalityBench Slop indexChat & Writing27.739.1+11.5
Artificial Analysis Intelligence Index v4.3.2Indexes56.045.1−10.9
Design Arena Overall Code CategoriesCode13491289−10.0
DAYJOB: TaxFinance & Accounting10.81.2−9.6
VCT (SecureBio)Science52.542.9−9.6
Artificial Analysis Progress Index v1Indexes351324−9.4
Artificial Analysis Strategy & Ops IndexIndexes58.149.2−8.8
VulcanBench Frontier v4 (functional pass@1)Code100.091.3−8.7
GDP.pdf (Artificial Analysis, All-pass)Professional Work25.817.2−8.6
LMArena Agent Arena Confirmed SuccessAgents19.52.3−8.6
DAXBench v2Code86.277.7−8.5
CursorBench 4.0Code55.547.7−7.8
DeepsecBench — Revalidated V1 (233 issues)Code37.530.0−7.5
CritPt (avg@5)Science31.424.0−7.4
NoCoT-Bench NCRI 15.2General QA95.181.6−7.4
LAB-Bench LitQA2 (SecureBio)Science55.849.0−6.8
Excel Modeling Benchmark v1.1Finance & Accounting75.969.2−6.7
Visa Agentic Commerce BenchmarkAgents64.457.8−6.7
LAB-Bench ProtocolQA (SecureBio)Science88.381.9−6.4
Harvey's Legal Agent BenchmarkLegal2.99.2+6.3
Agents' Last Exam v1.1 ScoreAgents60.466.6+6.2
MMMU ProMultimodal85.879.7−6.1
ARC-AGI-2General QA83.978.1−5.8
Epoch Capabilities Index (ECI)Indexes166159−5.6
LMArena Agent Arena Steering BurdenAgents14.54.0−5.2
APEX-Accounting 1.1Finance & Accounting64.559.3−5.2
LMArena Agent Arena Bash ReliabilityAgents12.72.4−5.2
SciCode 1.0.1 (Artificial Analysis)Code61.055.9−5.1
MBCT (SecureBio)Science58.853.9−4.9
WMDP-Bio (SecureBio)Science92.687.9−4.7
Vibe Code Bench v1.1Code92.487.7−4.7
RedlineBenchLegal40.344.9+4.6
VIALS-AA (pass@3)Multimodal70.866.5−4.3
APEX-Agents 1.1Professional Work83.779.4−4.3
SimpleQA Verified (Epoch anti-abstention)General QA46.542.5−4.0
AA-PersonalityBench Organization indexChat & Writing84.888.6+3.8
NoCoT-Bench NCKI (kspine v3)General QA107102−3.8
HLAB-Vals-1.1Legal10.414.2+3.8
AutomationBench-AA 1.0.6Agents71.868.5−3.3
LiveBench 2.0Agents82.579.2−3.3
o11y-bench 1.1 Pass^3Agents84.481.3−3.1
AA-LCR v1.1Search79.782.7+3.0
GDPxProfessional Work8.25.4−2.8
LAB-Bench CloningScenarios (SecureBio)Science97.094.8−2.2
FrontierCode 1.1 Main (Artificial Analysis)Code45.843.7−2.2
ARC-AGI-1 Semi-PrivateGeneral QA96.594.5−2.0
DualEntry Accounting AI BenchmarkFinance & Accounting80.282.2+2.0
Access to Justice BenchmarkLegal74.973.1−1.8
Ruby LLM Benchmarks — Program FixerCode73.872.0−1.8
PubMedQA (SecureBio)Health & Medicine81.380.0−1.3
VibeVideoBench (overall quality, no video = 0)Code3.93.9+0.6
OTIS Mock AIME 2024-2025Math100.099.4−0.6
LMArena Agent Arena Tool HallucinationAgents0.90.3−0.3
MMLU-Bio (SecureBio)Science91.090.7−0.3
GPQA-Bio (SecureBio)Science87.687.8+0.2

Claude Sonnet 5.5vsDeepSeek V4.1 Flash 1017normalised score marginsData Nov 15, 2026Data: Artificial Analysis, SecureBio, Surge AI and 36 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…