Atlas

Compare Models

DeepSeek V4.1 Flash vs GPT-5.5

GPT-5.5 leads by 0.6 capability points (≈95% interval −1.3 to +2.6): statistically tied · P(higher) ≈ 74%

Wins 84 of 151 shared benchmarks (64 lost, 3 tied)

Capability vs. relative cost
DeepSeek V4.1 FlashGPT-5.5Closed 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-5.5capability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins64 – 84 · 151 shared, 3 tied

Comparing matching setups only (40 benchmarks excluded) · suites counted once
← DeepSeek V4.1 FlashGPT-5.5 →

DeepSeek V4.1 FlashvsGPT-5.5wins by categoryData Nov 15, 2026Data: Vals AI, Mercor, benchmarks.bio and 33 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is GPT-5.5 minus DeepSeek V4.1 Flash.

KernelBench Hard Normalized (RTX PRO 6000)Code69.70.0−69.7
FrontierMath Tier 4 (v2) - PrivateMath26.872.5+45.7
SWE-bench Verified Extended (Mercor held-out)Code82.739.3−43.4
Surface Evolver BenchAgents46.388.1+41.9
KernelBench Mega Normalized (RTX PRO 6000)Code48.212.2−36.0
LLM Creative Story WritingChat & Writing-0.52.4+33.0
RiemannBenchMath24.055.2+31.2
BrokenArXiv 06/2026Math39.469.4+30.1
ArXivMath 06/2026Math57.683.6+26.0
A-Fantasia (Valid Attempts with Retries)Games59.334.7+24.7
ArXivMath 04/2026Math45.068.8+23.8
AutomationBench-AA 1.0.6Agents68.947.3−21.6
BrokenArXiv 05/2026Math28.550.0+21.5
SWE-bench MultilingualCode98.277.7−20.5
SimpleQA Verified (Epoch anti-abstention)General QA42.963.0+20.1
TaxCalcBench TY25Finance & Accounting4.024.0+20.0
APEX-Agents 1.1Professional Work50.570.3+19.8
SWE Atlas Codebase QnA Extended (Mercor held-out)Code36.717.3−19.4
ChartographyMultimodal11.631.0+19.4
ArXivMath 05/2026Math58.377.5+19.2
BuseyBench SVGMultimodal4.16.0+19.1
MMMU-Pro Extended (Mercor held-out)Multimodal25.043.8+18.8
LLM Chess EloGames10871532+18.3
FrontierMath Tiers 1-3 (v2) - PrivateMath67.485.3+17.9
YC-BenchGames16572361206045−15.8
AA-Briefcase EloAgents14241158−15.4
Creative Writing v3 EloChat & Writing15401845+15.4
APEX-SWECode51.936.6−15.3
AA-AnalystAgent (pass^5)Professional Work35.050.0+15.0
Chess Puzzles (task version 2.0)Games48.063.0+15.0
Vibe Code Bench v1.1Code84.769.8−14.9
Apex (MathArena)Math65.680.2+14.6
FoodTruck BenchGames5017961408+14.4
GBENCH One-shot Coding GScoreGames48.062.2+14.3
CharXiv Extended (Mercor held-out)Multimodal68.982.7+13.8
MedXpertQA MM (200-question set)Health & Medicine63.576.7+13.2
Mystery Game PuzzlesGames43.056.0+13.0
AA-Omniscience IndexIndexes-5.320.5+12.9
GDPval-AA v2Professional Work16321372−12.8
DAXBench v2Code75.588.3+12.8
BioSecBench-Function V1 Clean RerunsScience40.828.3−12.5
ACCoRDGeneral QA68.180.5+12.4
MedXpertQA MM Extended (Mercor held-out)Health & Medicine48.460.8+12.4
Conceptual Reasoning IndexIndexes58.467.3+12.2
Terminal-Bench 4.0 (Artificial Analysis)Code26.814.6−12.1
ARC-AGI-2General QA72.985.0+12.1
SkillsBench without SkillsAgents61.749.9−11.8
CritPt (avg@5)Science14.325.4+11.1
GDPval-AA v2.1Professional Work16001357−10.7
SpatialBench V0 VerifiedScience75.465.2−10.1
SUPERChemScience56.766.8+10.1
Furniture Assembly BenchmarkMultimodal34.244.2+10.0
BullshitBench v2 LeaderboardSafety57.047.0−10.0
ALE-BenchCode18292108+9.9
DeepSWE v1.1Code74.264.4−9.8
EnterpriseBench: CoreCraftProfessional Work41.551.3+9.8
Design Arena Overall Code CategoriesCode13161257−9.8
LiveBench 2.0Agents78.068.7−9.3
Creative Writing v3 Rubric ScoreChat & Writing14.115.9+9.2
Humanity's Last Exam DiamondGeneral QA19.127.9+8.8
AA-LCR v1.1Search55.364.0+8.7
TxBench Preclinical Pharmacology V1 Clean RerunsScience46.755.3+8.7
GameBench 2Games68.459.9−8.5
SWE Atlas - Codebase QnACode53.545.2−8.3
MedCodeHealth & Medicine41.249.1+7.9
DeepResearch Bench II (Mercor 130-task set)Search53.245.4−7.8
LLM Debate BenchmarkChat & Writing15821528−7.8
SkillsBench with SkillsAgents69.862.2−7.6
Excel Modeling BenchmarkFinance & Accounting57.264.5+7.3
LMCAGeneral QA47.054.3+7.3
Humanity's Last ExamGeneral QA36.843.8+7.0
ScBench-Long V1 Clean RerunsScience14.821.6+6.8
Senior SWE-BenchCode22.729.5+6.8
TxBench Antibody Discovery V1 (100 evaluations)Science39.733.0−6.7
Extended NYT Connections 940 (Quadratic v1)Games89.696.2+6.6
Finance Agent v2 - PrecedentsFinance & Accounting28.221.9−6.3
CyberGymCybersecurity88.181.8−6.3
GDP.pdfProfessional Work19.826.0+6.2
DTBenchGeneral QA89.996.0+6.1
SpatialBench-Long V1 Clean RerunsScience26.232.1+5.9
MMMU ProMultimodal65.671.4+5.8
TxBench Antibody Discovery V0Science39.033.6−5.4
Vision Arena OverallMultimodal12631283+5.2
LMArena Agent Arena SteerabilityAgents-3.36.9+5.1
Medical Long Context Reasoning (MLCR-AA)Health & Medicine22.817.8−5.0
ScBench-Long V0 VerifiedScience27.322.7−4.5
CharXiv (descriptive and reasoning combined)Multimodal90.194.6+4.5
BioSecBench-Surveillance V1 Clean RerunsScience51.547.1−4.4
Code2Video-AA EloCode10531000−4.3
SciCode 1.0.1 (Artificial Analysis)Code51.956.1+4.3
Longform Creative WritingChat & Writing74.078.2+4.2
SWE-bench VerifiedCode82.978.8−4.1
Finance Agent v2 - General Quantitative AnalysisFinance & Accounting69.873.7+3.9
SAGEMath47.951.5+3.7
Finance Agent v2 - AdjustmentsFinance & Accounting49.245.7−3.5
Public Benefits Bench v1.1Professional Work64.360.9−3.4
LegalBenchLegal83.386.5+3.2
Finance Agent v2 - Market AnalysisFinance & Accounting63.460.3−3.1
SRE Bench — Fully SolvedCode0.83.8+3.1
GeneBench-Pro Public Release SubsetScience11.08.0−3.0
Epoch Capabilities Index (ECI)Indexes156160+2.9
Harvey's Legal Agent BenchmarkLegal6.73.8−2.9
APEX-1 (Mercor, mean score)Professional Work65.568.4+2.9
EpiBench V1 Clean RerunsScience25.528.3+2.8
EpiBenchScience42.145.0+2.8
AdvancedIFChat & Writing84.081.3−2.7
Finance Agent v2 - Disclosure AnalysisFinance & Accounting61.358.9−2.4
Finance Agent v2 - ComparablesFinance & Accounting46.143.7−2.4
ScBench V1 Clean RerunsScience54.757.1+2.4
SimpleBenchGeneral QA66.769.0+2.3
BioSecBench-Function V0Science38.736.5−2.2
GDPval (Mercor 213-task rubric score)Professional Work86.784.6−2.1
Tax Agent BenchFinance & Accounting62.560.5−2.0
Terminal-Bench 2.1Code74.576.4+1.9
Finance Agent v2 - General Qualitative AnalysisFinance & Accounting67.465.6−1.8
AIME 2025Math98.3100.0+1.7
BrowseComp (Mercor 130-question subset)Search79.477.7−1.7
OTIS Mock AIME 2024-2025Math98.3100.0+1.7
SpatialBench-Long V0 VerifiedScience34.933.3−1.5
Finance Agent v2 - All PassFinance & Accounting41.139.6−1.5
MineBench — Bradley-Terry RatingGames18221846+1.5
MedScribeHealth & Medicine85.586.9+1.4
LMArena Agent Arena Net ImprovementAgents3.94.2+1.3
LMArena Text Arena EloChat & Writing14751481+1.1
ITBench-AAAgents46.945.8−1.1
Artificial Analysis Intelligence Index v4.3.2Indexes39.538.4−1.1
SciCode (main problems)Science11.212.3+1.1
Legal Research BenchLegal41.340.4−1.0
Artificial Analysis Intelligence Index v4.3Indexes39.538.6−0.9
TxBench Oligo Discovery V1 (113 evaluations)Science34.833.9−0.9
SpatialBench V1 Clean RerunsScience63.264.1+0.9
LMArena Agent Arena Bash RecoveryAgents7.99.5+0.8
SpeciEval (Pooled English Runs)Safety97.498.2+0.8
SpeciEval (Refusal-Aware Retries, Historical)Safety97.498.2+0.8
MMLU ProGeneral QA88.288.9+0.7
VariantBench V1 Clean RerunsScience31.130.5−0.6
ARC-AGI-1 Semi-PrivateGeneral QA94.595.0+0.5
Code MigrationCode45.645.2−0.5
Vals Index v2Indexes57.957.4−0.4
Terminal-Bench 2.1 Extended (Mercor held-out)Code38.438.0−0.4
AA-LCR Extended (Mercor held-out)Search70.870.5−0.3
TxBench Oligo Discovery V0Science33.933.6−0.3
Finance Agent v2 - Earnings AnalysisFinance & Accounting72.973.2+0.3
ComplexConstraintsChat & Writing49.749.5−0.2
Finance Agent v2 - Financial ModelingFinance & Accounting23.122.9−0.1
APEX-AccountingFinance & Accounting48.148.2+0.1
Artificial Analysis Progress Index v1Indexes320320−0.1
ARC-AGI-3 (Semi-Private)General QA0.50.4−0.1
CadQueryEval (Corrected Geometry)Code76.076.00.0
OncologyBench Strict (25 tasks, October 9 rubric)Health & Medicine14.714.70.0
ProgramBenchCode0.50.50.0

DeepSeek V4.1 FlashvsGPT-5.5normalised score marginsData Nov 15, 2026Data: Vals AI, Mercor, benchmarks.bio and 33 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…