Atlas

Compare Models

Gemini 3.7 Flash vs Gemini 4 Argon

Gemini 4 Argon leads by 6.5 capability points (≈95% interval +4.3 to +8.8): a clear lead over Gemini 3.7 Flash · P(higher) ≈ 100%

Wins 164 of 194 shared benchmarks (19 lost, 11 tied)

Capability vs. relative cost
Gemini 3.7 FlashGemini 4 ArgonClosed 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.

Gemini 3.7 FlashvsGemini 4 Argoncapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins19 – 164 · 194 shared, 11 tied

Comparing matching setups only (26 benchmarks excluded) · suites counted once
← Gemini 3.7 FlashGemini 4 Argon →

Gemini 3.7 FlashvsGemini 4 Argonwins by categoryData Nov 15, 2026Data: Vals AI, Arena, Center for AI Safety and 38 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Gemini 4 Argon minus Gemini 3.7 Flash.

Terminal-Bench 4.0Code6.157.4+51.3
Concept Synth INDUCTION Challenge64 Round 1General QA17.265.6+48.4
TaxCalcBench TY25 (web search)Finance & Accounting26.074.0+48.0
BrowseComp Extended (Mercor held-out)Search25.572.0+46.5
Concept Synth INDUCTION Challenge100 Round 1General QA11.057.0+46.0
Furniture Assembly BenchmarkMultimodal26.772.5+45.8
FrontierMath Tier 4 (v2) - PrivateMath36.678.0+41.5
Terminal-Bench 4.0 (Artificial Analysis)Code13.655.1+41.4
ProofBench v1.1Math58.099.0+41.0
Harvey's Legal Agent Benchmark - Emerging Companies Venture CapitalLegal0.040.0+40.0
Harvey's Legal Agent Benchmark - Funds Asset ManagementLegal0.040.0+40.0
Harvey's Legal Agent Benchmark - Trusts Estates Private ClientLegal0.040.0+40.0
Terminal-Bench Science 0.1 (Vals)Science7.144.3+37.1
RiemannBenchMath39.276.0+36.8
FrontierSWE V2 Worst@5Code10.344.5+34.2
ALE-BenchCode18402805+34.1
CyberBench v1.1 OverallCybersecurity43.877.9+34.1
Code MigrationCode34.868.2+33.4
FrontierSWE V2 Best@5Code31.264.3+33.1
IOI (Vals v2)Code67.8100.0+32.2
ChartographyMultimodal40.471.6+31.2
EnigmaEvalGeneral QA33.964.6+30.7
Agent Red Teaming k=100 (CAIS)Safety38.510.2+28.3
Mystery Game PuzzlesGames37.064.0+27.0
MineBench — Bradley-Terry RatingGames17652195+26.0
AA-LCR Extended (Mercor held-out)Search48.974.7+25.8
LMArena Agent Arena Net Improvement (Steering Burden)Agents-2.210.7+24.4
TextQuests Game ProgressGames52.977.3+24.4
Bug Hunt BenchCode15.238.7+23.5
HackerRank-BenchCode27.349.7+22.4
Tax Agent BenchFinance & Accounting57.779.0+21.4
VIALSMultimodal26.547.2+20.7
LLM Chess EloGames11071609+20.6
AA-Briefcase v1.1 EloAgents11151485+20.3
Legal Research BenchLegal34.654.8+20.2
Harvey's Legal Agent Benchmark - Antitrust CompetitionLegal0.020.0+20.0
Harvey's Legal Agent Benchmark - Banking FinanceLegal20.00.0−20.0
Harvey's Legal Agent Benchmark - Capital MarketsLegal20.040.0+20.0
Harvey's Legal Agent Benchmark - Corporate M&ALegal0.020.0+20.0
Harvey's Legal Agent Benchmark - Data Privacy CybersecurityLegal0.020.0+20.0
Harvey's Legal Agent Benchmark - Energy Natural ResourcesLegal40.020.0−20.0
Harvey's Legal Agent Benchmark - ImmigrationLegal0.020.0+20.0
Harvey's Legal Agent Benchmark - InsuranceLegal20.00.0−20.0
Harvey's Legal Agent Benchmark - Intellectual PropertyLegal0.020.0+20.0
Harvey's Legal Agent Benchmark - International Trade SanctionsLegal0.020.0+20.0
Terminal-Bench 2.1Code77.557.7−19.9
SpatialViz-BenchMultimodal75.394.2+18.9
VariantBench V1 Clean RerunsScience31.450.3+18.9
Ramp Accounting BenchFinance & Accounting42.361.1+18.8
OncologyBench Strict (25 tasks, October 9 rubric)Health & Medicine14.733.3+18.7
GDPval Extended (Mercor held-out rubric score)Professional Work57.876.4+18.6
RSI-Exam 0.1 (Full 88)Agents0.30.5+18.5
BioSecBench-Surveillance V1 Clean RerunsScience41.559.8+18.3
MirrorCode (ML, +Private, 2L)Code41.158.9+17.8
LLM Creative Story WritingChat & Writing-0.70.8+17.0
LiveBench 2.0Agents67.284.1+16.9
Arena Code WebDev — Content Creation ToolsCode15741676+16.1
Arena Code WebDev — ReactCode15841704+16.1
Finance Agent v2 - PrecedentsFinance & Accounting33.849.8+16.0
DAYJOB: FinanceFinance & Accounting2.818.8+16.0
HPCT Refusal (CAIS)Safety13.029.0+16.0
Humanity's Sixth SenseMultimodal39.855.4+15.6
AutomationBench-AA 1.0.6Agents62.077.5+15.5
GBENCH One-shot Coding GScoreGames52.667.9+15.3
Toloka Arena (v3)Agents58.073.0+15.0
HieroglyphBenchMultimodal55.069.7+14.7
FrontierMath Tiers 1-3 (v2) - PrivateMath71.686.0+14.4
APEX-Accounting 1.1Finance & Accounting57.271.5+14.3
Arena Code WebDev — SimulationsCode16181742+14.1
Arena Code WebDev — GamingCode16381763+14.0
NoCoT-Bench NCRI 15.2General QA126151+13.8
Artificial Analysis Intelligence Index v4.3.2Indexes39.152.6+13.5
LMArena Agent Arena Praise vs ComplaintAgents-2.924.1+13.5
LMCAGeneral QA50.463.7+13.3
TxBench Oligo Discovery V1 (113 evaluations)Science37.851.0+13.3
ChessBench EloGames12651485+13.1
TxBench Antibody Discovery V1 (100 evaluations)Science50.763.7+13.0
Senior SWE-BenchCode14.727.7+13.0
CritPt (avg@5)Science14.327.1+12.9
Arena Code WebDev — Reference-Based DesignCode16081703+12.5
DeepSWE v1.1Code65.577.9+12.4
Vals RSI Index v1.1 (four tasks)Agents18.330.6+12.3
Arena Code WebDev — FrontendCode15921688+12.3
Remote Labor Index (RLI)Professional Work5.017.1+12.1
Eval Connections One-Shot with TrapsGames80.092.0+12.0
Finance Agent v2 - General Qualitative AnalysisFinance & Accounting65.377.0+11.7
Wiz Cyber Model Arena v3.9Cybersecurity65.677.0+11.4
Humanity's Last Exam DiamondGeneral QA31.242.5+11.3
AA-AnalystAgent (pass^5)Professional Work60.048.8−11.2
ComplexConstraintsChat & Writing42.453.3+10.9
Arena Code WebDev — FullstackCode16291671+10.9
DAYJOB: HealthcareHealth & Medicine0.010.8+10.8
Conceptual Reasoning IndexIndexes64.672.4+10.8
BioSecBench-Function V1 Clean RerunsScience33.744.4+10.8
MetagenomicsBench V1Science31.742.3+10.7
LMArena Agent Arena Steering BurdenAgents-7.313.2+10.3
GDPval-AA v2.1Professional Work14041634+10.1
Humanity's Last ExamGeneral QA45.655.6+10.0
SpatialBench V1 Clean RerunsScience59.169.0+9.9
Arena Code WebDev — Brand & MarketingCode15531620+9.7
ScBench-Long V1 Clean RerunsScience18.828.4+9.7
HPCT (SecureBio)Science64.774.2+9.5
ChessBench CoherenceGames83.292.6+9.4
TxBench Preclinical Pharmacology V1 Clean RerunsScience55.364.7+9.3
Arena Code WebDev — HTMLCode15931660+9.3
BuseyBench SVGMultimodal6.67.6+9.2
LAB-Bench ProtocolQA (SecureBio)Science86.895.9+9.1
VCT (SecureBio)Science49.358.4+9.1
ERQAMultimodal77.786.5+8.8
ScBench V1 Clean RerunsScience50.459.1+8.7
VCT v2 (SecureBio)Science48.657.3+8.7
Artificial Analysis Progress Index v1Indexes317342+8.6
LMArena Agent Arena Confirmed SuccessAgents-3.513.2+8.4
Finance Agent v2 - All PassFinance & Accounting47.555.8+8.3
MindCube TinyMultimodal88.496.7+8.3
MBCT (SecureBio)Science55.063.2+8.2
Bio Capabilities Index (SecureBio)Indexes162174+8.2
APEX-Agents 1.1Professional Work79.487.4+8.0
LMArena Agent Arena Bash RecoveryAgents-5.4-21.4−8.0
AA-Omniscience IndexIndexes26.542.4+7.9
SpeciEval (Pooled English Runs)Safety89.081.3−7.8
GPQA-Bio (SecureBio)Science81.689.3+7.7
MASKSafety53.761.0+7.3
APEX-SWECode42.049.2+7.2
Chess Puzzles (task version 2.0)Games66.059.0−7.0
EpiBench V1 Clean RerunsScience22.028.9+6.9
SimpleQA Verified (Epoch anti-abstention)General QA69.276.0+6.8
Finance Agent v2 - Disclosure AnalysisFinance & Accounting67.774.5+6.8
ARC-AGI-2General QA84.691.3+6.7
Epoch Capabilities Index (ECI)Indexes158167+6.5
IntPhys 2Multimodal68.274.6+6.4
PCT Political Inconsistency (CAIS)Safety45.138.7+6.4
BioSecBench-Refusal V1 Clean RerunsSafety54.848.7−6.1
VCT-Refusal (CAIS)Safety4.810.7+5.9
TextQuests Average Cumulative HarmSafety24.426.1−5.9
Finance Agent v2 - AdjustmentsFinance & Accounting54.760.5+5.8
LMArena Agent Arena Bash ReliabilityAgents6.2-5.4−5.8
LMArena Text Arena EloChat & Writing14891522+5.8
Finance Agent v2 - Earnings AnalysisFinance & Accounting79.184.8+5.8
Finance Agent v2 - Financial ModelingFinance & Accounting24.229.8+5.6
Harvey HLAB — Criteria Pass RateLegal88.694.0+5.4
MedCodeHealth & Medicine53.458.8+5.4
Arena Code WebDev — Data & AnalyticsCode15771612+5.3
Finance Agent v2 - Market AnalysisFinance & Accounting74.579.7+5.2
NoCoT-Bench NCKI (kspine v3)General QA121128+5.0
Arena Alignment IndexSafety74.479.4+5.0
HardGeoBench Country AccuracyMultimodal67.562.5−5.0
GDP.pdfProfessional Work23.828.4+4.6
SAGEMath49.253.6+4.4
FrontierCode 1.1 Extended - ScoreCode56.360.5+4.2
Arena Code WebDev — Consumer ProductCode15741607+4.2
MedXpertQA MM Extended (Mercor held-out)Health & Medicine62.766.7+4.0
Excel Modeling BenchmarkFinance & Accounting71.375.2+3.9
FrontierCode 1.1 Extended - Pass RateCode62.366.1+3.8
HardGeoBench <=10miMultimodal46.342.5−3.7
Design Arena Overall Code CategoriesCode13041326+3.6
MedScribeHealth & Medicine83.987.4+3.5
MMMU ProMultimodal84.988.3+3.5
AA-LCR v1.1Search83.079.7−3.3
WCB (SecureBio)Science58.061.1+3.1
ARC-AGI-1 Semi-PrivateGeneral QA95.598.5+3.0
BioTIER RefusalSafety84.487.2+2.8
ChessBench AccuracyGames79.281.8+2.8
OTIS Mock AIME 2024-2025Math97.2100.0+2.8
ProgramBenchCode0.02.5+2.5
SciCode 1.0.1 (Artificial Analysis)Code59.861.8+2.0
Creative Writing v3 EloChat & Writing17201758+1.9
Creative Writing v3 Rubric ScoreChat & Writing15.215.6+1.7
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine12.213.9+1.7
Machiavelli Deception (CAIS)Safety85.585.0+1.5
HardGeoBench <=100miMultimodal56.355.0−1.2
SpatialBench-Long V1 Clean RerunsScience29.828.6−1.2
LegalBenchLegal87.388.3+1.0
DTBenchGeneral QA96.897.6+0.8
GDPval (Mercor 213-task rubric score)Professional Work85.285.9+0.7
MMLU-Bio (SecureBio)Science91.492.1+0.7
Finance Agent v2 - ComparablesFinance & Accounting50.351.0+0.7
LAB-Bench CloningScenarios (SecureBio)Science99.4100.0+0.6
HardGeoBench Avg DistanceMultimodal895880+0.5
ACCoRDGeneral QA77.077.3+0.3
DistressBenchSafety69.068.6−0.3
Finance Agent v2 - General Quantitative AnalysisFinance & Accounting81.881.5−0.3
LMArena Agent Arena Tool HallucinationAgents0.70.5−0.1
Harvey's Legal Agent Benchmark - Arbitration International Dispute ResolutionLegal0.00.00.0
Harvey's Legal Agent Benchmark - Bankruptcy RestructuringLegal0.00.00.0
Harvey's Legal Agent Benchmark - Corporate GovernanceLegal20.020.00.0
Harvey's Legal Agent Benchmark - Employment LaborLegal0.00.00.0
Harvey's Legal Agent Benchmark - Environmental ESGLegal0.00.00.0
Harvey's Legal Agent Benchmark - Healthcare Life SciencesLegal20.020.00.0
Harvey's Legal Agent Benchmark - Litigation Dispute ResolutionLegal0.00.00.0
Harvey's Legal Agent Benchmark - Real EstateLegal0.00.00.0
Harvey's Legal Agent Benchmark - Structured Finance SecuritizationLegal0.00.00.0
Harvey's Legal Agent Benchmark - TaxLegal0.00.00.0
Harvey's Legal Agent Benchmark - White Collar Defense InvestigationsLegal0.00.00.0

Gemini 3.7 FlashvsGemini 4 Argonnormalised score marginsData Nov 15, 2026Data: Vals AI, Arena, Center for AI Safety and 38 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…