Atlas

Compare Models

DeepSeek V4.1 Flash vs GPT-6.1 Sol

GPT-6.1 Sol leads by 6.9 capability points (≈95% interval +4.6 to +9.2): a clear lead over DeepSeek V4.1 Flash · P(higher) ≈ 100%

Wins 162 of 201 shared benchmarks (37 lost, 2 tied)

Capability vs. relative cost
DeepSeek V4.1 FlashGPT-6.1 SolClosed 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 FlashvsGPT-6.1 Solcapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins37 – 162 · 201 shared, 2 tied

Comparing matching setups only (60 benchmarks excluded) · suites counted once
← DeepSeek V4.1 FlashGPT-6.1 Sol →

DeepSeek V4.1 FlashvsGPT-6.1 Solwins by categoryData Nov 15, 2026Data: Vals AI, Artificial Analysis, Mercor and 41 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is GPT-6.1 Sol minus DeepSeek V4.1 Flash.

ARC-AGI-3 (Semi-Private)General QA2.796.2+93.5
Concept Synth INDUCTION Challenge100 Round 1General QA11.090.0+79.0
Concept Synth INDUCTION Challenge64 Round 1General QA17.295.3+78.1
FrontierMath Tier 4 (v2) - PrivateMath26.8100.0+73.2
Design Creativity BenchCode43.663.7+63.9
BrokenArXiv 05/2026Math28.590.0+61.5
BrokenArXiv 06/2026Math39.4100.0+60.7
IOI (Vals v2)Code40.396.9+56.6
Terminal-Bench Science 0.1 (Vals)Science0.052.9+52.9
AA-PersonalityBench Informativeness indexChat & Writing79.127.7−51.9
SRE Bench — Fully SolvedCode0.850.8+50.0
Terminal-Bench-Science 0.1Science9.058.1+49.0
Furniture Assembly BenchmarkMultimodal34.280.0+45.8
ProofBench v1.1Math54.099.0+45.0
AA-PersonalityBench Conviction indexChat & Writing96.254.3−42.3
ALE-BenchCode18292990+41.1
FrontierCUA pass@1Agents6.746.7+40.0
Humanity's Last Exam DiamondGeneral QA19.158.8+39.7
BuseyBench SVGMultimodal4.18.0+38.6
SWE-bench Verified Extended (Mercor held-out)Code82.745.3−37.4
Mystery Game PuzzlesGames43.080.0+37.0
ArXivMath 06/2026Math57.694.4+36.8
ArXivMath 05/2026Math58.395.0+36.7
VIALSMultimodal12.649.1+36.5
MMMU-Pro Extended (Mercor held-out)Multimodal25.061.3+36.3
CyberBench v1.1 OverallCybersecurity73.739.3−34.4
KernelBench Mega Normalized (RTX PRO 6000)Code48.214.1−34.2
TaxCalcBench TY25Finance & Accounting4.038.0+34.0
AA-PersonalityBench Transparency indexChat & Writing60.092.4+32.7
WeirdML v3Science5.838.3+32.4
SAFE-Teen 106 worst branch: Health Advice (no prompt)Safety41.18.8+32.3
SAFE-Teen 106 worst branch: Relationships (no prompt)Safety51.119.1+32.0
GBENCH One-shot Coding GScoreGames48.079.0+31.0
SimpleQA Verified (Epoch anti-abstention)General QA42.973.9+31.0
AA-PersonalityBench Imagination indexChat & Writing31.561.9+30.8
Vals Science IndexIndexes29.659.6+30.0
SAFE-Teen 106 worst branch: Health Advice (under-18 prompt)Safety37.07.7+29.3
Terminal-Bench 4.0 (Artificial Analysis)Code26.856.1+29.3
OpenWorldEngineer (Mean@8)Code12.441.2+28.8
RareDiseaseBench V1 (35 evaluations)Science0.027.6+27.6
LMArena Agent Arena Work Net Improvement (Steering Burden)Agents2.513.6+26.7
LMArena Agent Arena Code Net Improvement (Steering Burden)Agents2.714.5+26.5
FrontierMath Tiers 1-3 (v2) - PrivateMath67.493.7+26.3
BioSecBench-Refusal V1 Clean RerunsSafety11.337.6+26.3
Code2Video-AA EloCode10531369+25.7
SAFE-Teen 106: Health Advice (no prompt)Safety30.24.7+25.5
MysteryMechanismScience21.246.4+25.2
YC-BenchGames1657236936267−25.2
AA-PersonalityBench Warmth indexChat & Writing41.065.7+25.0
LMArena Agent Arena Chat Net Improvement (Steering Burden)Agents1.611.6+24.5
Decision-Bench v1.1Professional Work25.349.1+23.8
GameBench 2Games68.492.0+23.5
AA-Omniscience IndexIndexes-5.341.5+23.4
AA-PersonalityBench Slop indexChat & Writing37.214.3−23.1
APEX-Agents 1.1Professional Work50.573.2+22.7
Terminal-Bench 4.0Code31.253.4+22.2
MedXpertQA MM (200-question set)Health & Medicine63.585.7+22.2
GeneBench-Pro Public Release SubsetScience11.033.0+22.0
LLM Chess EloGames10871612+21.5
ARC-AGI-2General QA72.994.2+21.3
SUPERChemScience56.777.8+21.1
Wiz Cyber Model Arena v3.9Cybersecurity50.270.7+20.5
SAFE-Teen 106: Health Advice (under-18 prompt)Safety23.63.3+20.3
Chess Puzzles (task version 2.0)Games48.068.0+20.0
TissueMicroscopyBench V1 (49 evaluations)Science2.722.4+19.7
Conceptual Reasoning IndexIndexes58.472.7+19.6
Code MigrationCode45.665.1+19.5
AA-PersonalityBench Organization indexChat & Writing10081.0−19.2
AutomationBench 1.0.6 (provider-reported)Agents54.836.1−18.7
Code Migration v1.1Code58.977.3+18.4
TxBench Oligo Discovery V1 (113 evaluations)Science34.852.5+17.7
CritPt (avg@5)Science14.331.7+17.4
DAXBench v2Code75.592.6+17.0
SAFE-Teen 106: Relationships (no prompt)Safety27.110.2+16.9
MedXpertQA MM Extended (Mercor held-out)Health & Medicine48.464.7+16.3
SimpleBenchGeneral QA66.782.9+16.2
CharXiv Extended (Mercor held-out)Multimodal68.985.1+16.2
SAFE-Teen 106 worst branch: Urgent Safety (no prompt)Safety31.415.4+16.1
BrowseComp (Mercor 130-question subset)Search79.495.4+16.0
LMCAGeneral QA47.062.8+15.8
HealthBench ProfessionalHealth & Medicine52.968.3+15.4
SAFE-Teen 106 worst branch: Relationships (under-18 prompt)Safety19.84.5+15.3
GDP.pdfProfessional Work19.835.0+15.2
AA-AnalystAgent (pass^5)Professional Work35.050.0+15.0
VariantBench V1 Clean RerunsScience31.146.0+15.0
SAFE-Teen 106 worst branch: Urgent Safety (under-18 prompt)Safety27.212.7+14.5
Humanity's Last ExamGeneral QA36.851.3+14.5
TreasuryQA BenchProfessional Work37.151.4+14.3
APEX-Accounting 1.1Finance & Accounting50.164.3+14.2
SpatialBench-Long V1 Clean RerunsScience26.239.9+13.7
LMArena Agent Arena Work Praise vs ComplaintAgents3.730.9+13.6
Excel Modeling BenchmarkFinance & Accounting57.270.8+13.6
LMArena Agent Arena Code Praise vs ComplaintAgents3.029.9+13.4
TxBench-Antibody-Development V1 (60 evaluations)Science14.427.8+13.3
SWE-bench MultilingualCode98.284.9−13.3
Arena Alignment IndexSafety74.887.9+13.2
TxBench Preclinical Pharmacology V1 Clean RerunsScience46.759.7+13.0
ACCoRDGeneral QA68.180.9+12.8
Artificial Analysis Intelligence Index v4.3.2Indexes39.551.8+12.4
ScBench-Long V1 Clean RerunsScience14.826.7+11.9
BioMysteryBench (Vals Terminus 2)Science67.879.6+11.9
AA-PersonalityBench Expressiveness indexChat & Writing61.950.5−11.5
BigFinanceBench private holdout (final-answer accuracy)Finance & Accounting40.652.0+11.4
Medical Long Context Reasoning (MLCR-AA)Health & Medicine22.833.9+11.1
LMArena Agent Arena Chat Praise vs ComplaintAgents3.125.2+11.1
Bedside BenchHealth & Medicine82.993.9+11.0
Hebbia Financial Services BenchmarkFinance & Accounting63.173.5+10.4
Epoch Capabilities Index (ECI)Indexes156169+10.0
TxBench Antibody Discovery V1 (100 evaluations)Science39.749.7+10.0
Artificial Analysis Progress Index v1Indexes320348+9.9
OncologyBench Strict (25 tasks, October 9 rubric)Health & Medicine14.724.0+9.3
τ³-Banking (Fireworks, banking_knowledge)Agents35.744.7+8.9
ScBench V1 Clean RerunsScience54.763.6+8.9
Real-SWE 1.1Code38.447.2+8.8
EnterpriseBench: CoreCraftProfessional Work41.550.3+8.8
BullshitBench v2 LeaderboardSafety57.065.0+8.0
SAFE-Teen 106: Urgent Safety (no prompt)Safety11.33.4+8.0
LMArena Agent Arena Code Steering BurdenAgents2.318.1+7.9
AA-Briefcase v1.1 EloAgents14151555+7.7
MetagenomicsBench V1Science38.045.7+7.7
MedCodeHealth & Medicine41.248.8+7.7
Vision Arena OverallMultimodal12631293+7.6
Terminal-Bench 2.1 Extended (Mercor held-out)Code38.445.8+7.4
LiveBench 2.0Agents78.085.4+7.4
BioSecBench-RespiratoryVirus-Surveillance V1 (110 evaluations)Science5.212.4+7.3
AML BenchFinance & Accounting63.470.6+7.2
Access to Justice BenchmarkLegal72.965.8−7.0
Epiq LAER CorporateBenchProfessional Work75.382.2+6.9
Artificial Analysis Coding Agent Index v2.0Indexes47.354.1+6.8
Westworld Finance Diligence BenchFinance & Accounting48.355.0+6.7
DTBenchGeneral QA89.996.5+6.7
Design Arena Overall Code CategoriesCode13161356+6.6
LMArena Agent Arena Work Steering BurdenAgents1.514.7+6.6
MacroscopeBenchCode72.078.5+6.5
BioSecBench-Structure V1 (58 evaluations)Science27.621.3−6.3
CharXiv (descriptive and reasoning combined)Multimodal90.196.0+5.9
Extended NYT Connections 940 (Quadratic v1)Games89.695.5+5.9
SAFE-Teen 106: Urgent Safety (under-18 prompt)Safety7.92.0+5.9
DistressBenchSafety79.874.2−5.6
Ramp Accounting BenchFinance & Accounting40.846.4+5.6
SWE Atlas Codebase QnA Extended (Mercor held-out)Code36.731.3−5.4
LMArena Agent Arena Chat Bash ReliabilityAgents-1.09.7+5.4
Harvey LAB-AA v1.1 (hallucination-gated)Legal1.76.9+5.3
τ³-Banking (Mercor run, bm25_grep retrieval)Agents37.843.0+5.2
Terminal-Bench 2.1Code90.685.4−5.2
BioSecBench-Function V1 Clean RerunsScience40.846.0+5.2
SAGE v1.1Math85.490.5+5.1
SWE Atlas - Codebase QnACode53.548.4−5.1
FrontierCode 1.1 Main (Artificial Analysis)Code44.049.1+5.1
APEX-SWECode51.946.9−5.0
HealthBenchHealth & Medicine61.956.9−5.0
Public Benefits Bench v1.1Professional Work64.359.3−4.9
LMArena Agent Arena Chat Steering BurdenAgents-0.59.3+4.9
LMArena Agent Arena Code Confirmed SuccessAgents6.615.6+4.5
SWE-bench VerifiedCode82.987.3+4.4
Harvey LAB: In-House Contracting (held-out)Legal0.34.7+4.3
Financial Coaching BenchFinance & Accounting72.868.5−4.3
Vibe Code Bench v1.1Code84.788.9+4.2
SpatialBench V1 Clean RerunsScience63.267.3+4.1
ARC-AGI-1 Semi-PrivateGeneral QA94.598.5+4.0
SciCode 1.0.1 (Artificial Analysis)Code51.955.8+3.9
LMArena Agent Arena Work Confirmed SuccessAgents5.713.5+3.9
LMArena Agent Arena Chat Confirmed SuccessAgents4.912.5+3.8
KernelBench CUDA Normalized (RTX PRO 6000)Code47.043.2−3.7
Finance Agent v2.1Finance & Accounting65.461.8−3.6
SciCode (main problems)Science11.214.6+3.4
Excel Modeling Benchmark v1.1Finance & Accounting67.871.1+3.3
SAFE-Teen 106: Relationships (under-18 prompt)Safety4.20.9+3.3
DualEntry Accounting AI BenchmarkFinance & Accounting84.281.2−3.0
SWE-bench Multimodal (Vals v1, 480 Tasks)Code46.543.5−2.9
Legal Research BenchLegal41.338.5−2.9
EpiBench V1 Clean RerunsScience25.528.3+2.8
MMLU ProGeneral QA88.291.0+2.8
Legacy-Bench (Fireworks, avg@3)Code57.755.0−2.7
DeepSWE v1.1Code72.675.2+2.6
ProgramBenchCode0.53.0+2.5
HealthBench (length-adjusted, Artificial Analysis)Health & Medicine60.958.5−2.4
AutomationBench-AA 1.0.6Agents68.966.6−2.3
Immigration BenchLegal65.667.8+2.2
SheetBench-50Finance & Accounting58.460.5+2.1
AIME 2025Math98.3100.0+1.7
OTIS Mock AIME 2024-2025Math98.3100.0+1.7
LMArena Agent Arena Bash RecoveryAgents7.94.5−1.7
AdvancedIFChat & Writing84.082.4−1.6
Finance Agent v2Finance & Accounting53.552.0−1.4
SAGEMath47.946.5−1.3
Vals Finance IndexIndexes57.155.8−1.3
Harvey's Legal Agent BenchmarkLegal6.75.4−1.2
LMArena Text Arena EloChat & Writing14751482+1.2
LMArena Agent Arena Work Bash ReliabilityAgents0.12.4+1.1
MedScribeHealth & Medicine85.586.5+1.0
LMArena Agent Arena Code Bash ReliabilityAgents-0.21.4+0.8
BioSecBench-Surveillance V1 Clean RerunsScience51.550.8−0.7
GDPval-AA v2.1Professional Work16001607+0.3
Tax Agent BenchFinance & Accounting62.562.3−0.1
LMArena Agent Arena Work Tool HallucinationAgents0.50.7+0.1
LMArena Agent Arena Code Tool HallucinationAgents0.50.7+0.1
LMArena Agent Arena Chat Tool HallucinationAgents0.30.5+0.1
NeuroscienceBench V1 (91 evaluations)Science22.022.00.0
SpeciEval (Pooled English Runs)Safety97.497.40.0
SpeciEval (Refusal-Aware Retries, Historical)Safety97.497.40.0

DeepSeek V4.1 FlashvsGPT-6.1 Solnormalised score marginsData Nov 15, 2026Data: Vals AI, Artificial Analysis, Mercor and 41 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…