Atlas

Compare Models

Gemini 3.7 Flash vs GPT-5.6 Terra

GPT-5.6 Terra leads by 1.3 capability points (≈95% interval −0.6 to +3.3): statistically tied · P(higher) ≈ 90%

Gemini 3.7 Flash wins 106 of 209 shared benchmarks (86 lost, 17 tied)

Capability vs. relative cost
Gemini 3.7 FlashGPT-5.6 TerraClosed 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 FlashvsGPT-5.6 Terracapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins106 – 86 · 209 shared, 17 tied

Comparing matching setups only (63 benchmarks excluded) · suites counted once
← Gemini 3.7 FlashGPT-5.6 Terra →

Gemini 3.7 FlashvsGPT-5.6 Terrawins by categoryData Nov 15, 2026Data: Vals AI, Mercor, Artificial Analysis and 44 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is GPT-5.6 Terra minus Gemini 3.7 Flash.

SWE-bench Verified Extended (Mercor held-out)Code30.078.7+48.7
Olam Social Poker EloGames14681511+46.7
Harvey's Legal Agent Benchmark - Energy Natural ResourcesLegal40.00.0−40.0
MASKSafety53.793.3+39.6
HLE Extended (Mercor held-out)General QA31.670.7+39.1
Odysseys2Agents5.844.6+38.8
HieroglyphBenchMultimodal55.020.5−34.5
FrontierMath Tier 4 (v2) - PrivateMath36.670.7+34.1
DeepSWE v1.1Code65.535.1−30.4
ChessBench — Historical Output RepairGames70.740.7−30.0
CyberBench v1.1 OverallCybersecurity43.872.1+28.3
Gray Swan IPI Attack Success (k=15)Safety9.237.3−28.1
Furniture Assembly BenchmarkMultimodal26.754.2+27.5
RiemannBenchMath39.266.4+27.2
Shroombench (mushroom identification, top-1)Multimodal61.035.0−26.0
MindCube TinyMultimodal88.462.6−25.8
AA-LCR Extended (Mercor held-out)Search48.974.4+25.5
SM BenchSafety80.659.0−21.6
NoCoT-Bench NCRI 15.2General QA12689.0−20.4
Concept Synth INDUCTION Challenge64 Round 1General QA17.237.5+20.3
ChessBench EloGames1265925−20.2
Terminal-Bench 4.0Code6.126.3+20.2
Harvey's Legal Agent Benchmark - Banking FinanceLegal20.00.0−20.0
Harvey's Legal Agent Benchmark - Capital MarketsLegal20.00.0−20.0
Harvey's Legal Agent Benchmark - Corporate GovernanceLegal20.00.0−20.0
Harvey's Legal Agent Benchmark - Healthcare Life SciencesLegal20.00.0−20.0
Harvey's Legal Agent Benchmark - InsuranceLegal20.00.0−20.0
Harvey's Legal Agent Benchmark - Structured Finance SecuritizationLegal0.020.0+20.0
IOI (Vals v2)Code67.887.6+19.8
FACTS Benchmark SuiteGeneral QA71.252.0−19.2
TxBench Antibody Discovery V0Science50.031.0−19.0
Chess Puzzles (task version 2.0)Games66.047.0−19.0
BioTIER RefusalSafety84.465.6−18.8
Humanity's Sixth SenseMultimodal39.822.7−17.1
Medical Long Context Reasoning (MLCR-AA)Health & Medicine15.031.7+16.7
Agent Red Teaming k=100 (CAIS)Safety38.554.6−16.1
ProofBench v1.1Math58.074.0+16.0
CadQueryEval (Corrected Geometry)Code84.068.0−16.0
HyperBrowseCompSearch31.715.8−15.8
CritPt (avg@5)Science14.330.0+15.7
Machiavelli Deception (CAIS)Safety85.580.4+15.5
SpatialBench-Long V0 VerifiedScience22.737.9+15.2
SWE-bench VerifiedCode80.895.4+14.6
FrontierMath Tiers 1-3 (v2) - PrivateMath71.686.0+14.4
ERQAMultimodal77.763.6−14.1
Roboflow Vision Evals (October 2026, five tasks)Multimodal80.166.0−14.1
Concept Synth INDUCTION Challenge100 Round 1General QA11.025.0+14.0
Finance Agent v2 - Market AnalysisFinance & Accounting74.560.6−13.9
ChessBench CoherenceGames83.269.7−13.5
AA-Omniscience IndexIndexes26.50.1−13.2
Toloka Arena (v1)Agents54.641.4−13.2
Code MigrationCode34.847.8+13.0
Senior SWE-BenchCode14.727.4+12.7
TextQuests Game ProgressGames52.940.5−12.4
Terminal-Bench 4.0 (Artificial Analysis)Code13.61.5−12.1
NoCoT-Bench NCKI (kspine v3)General QA121104−12.1
BabaIsBench Stage 0: The IntroGames100.088.0−12.0
HPCT Refusal (CAIS)Safety13.025.0+12.0
EnterpriseOps-Gym-AAAgents50.438.5−11.9
SWE-bench MultilingualCode82.694.4+11.8
MedXpertQA MM Extended (Mercor held-out)Health & Medicine62.751.0−11.7
Roboflow Vision Evals (September 2026 protocol)Multimodal85.273.8−11.4
MMMU-Pro Extended (Mercor held-out)Multimodal36.447.4+11.0
Design Arena Overall Code CategoriesCode13041238−10.9
APEX-SWECode42.031.1−10.9
EnigmaEvalGeneral QA33.923.1−10.8
GBENCH One-shot Coding GScoreGames52.663.3+10.7
ScBench-Long V0 VerifiedScience24.234.9+10.6
GDPval Extended (Mercor held-out rubric score)Professional Work57.847.2−10.6
SpatialViz-BenchMultimodal75.365.2−10.1
Terminal-Bench 2.1Code85.875.7−10.1
MedCodeHealth & Medicine53.443.4−10.0
ALE-BenchCode18762154+9.8
LAB-Bench ProtocolQA (SecureBio)Science86.877.0−9.8
Humanity's Last ExamGeneral QA45.635.9−9.7
Finance Agent v2 - General Quantitative AnalysisFinance & Accounting81.872.7−9.1
SpeciEval (Refusal-Aware Retries, Historical)Safety89.098.1+9.0
SpeciEval (Pooled English Runs)Safety89.098.1+9.0
Finance Agent v2 - PrecedentsFinance & Accounting33.824.8−9.0
Harvey HLAB — Criteria Pass RateLegal88.679.8−8.8
MRCR v2 (8-needle, Context Arena) - 128K pointwiseGeneral QA88.680.2−8.4
Terminal-Bench 2.1 Extended (Mercor held-out)Code44.436.4−8.0
HealthBench Professional (Length-Adjusted)Health & Medicine48.456.3+7.9
SUPERChemScience72.864.9−7.9
Document ArenaProfessional Work14761468−7.9
ChessBench AccuracyGames79.271.9−7.8
LiveBench 2.0Agents67.275.0+7.8
MedXpertQA MM (200-question set)Health & Medicine83.575.7−7.8
LLM Chess EloGames11071293+7.6
APEX-Agents 1.1Professional Work79.471.8−7.6
Tax Agent BenchFinance & Accounting57.765.2+7.5
DistressBenchSafety69.076.4+7.5
TextQuests Average Cumulative HarmSafety24.422.2+7.3
Vision Arena OverallMultimodal12941266−7.0
SkillsBench with SkillsAgents65.958.9−7.0
IntPhys 2Multimodal68.261.3−6.9
Creative Writing v3 EloChat & Writing17201856+6.9
LAB-Bench CloningScenarios (SecureBio)Science99.492.5−6.9
Legal Research BenchLegal34.641.3+6.7
VCT v2 (SecureBio)Science48.641.9−6.7
BioSecBench-Function V0Science40.133.5−6.6
BrowseComp (Mercor 130-question subset)Search79.285.8+6.6
LMArena Agent Arena Net ImprovementAgents-1.30.6+6.4
ChartographyMultimodal40.434.0−6.4
TxBench Oligo Discovery V0Science39.533.3−6.2
SpatialBench V0 VerifiedScience64.370.4+6.1
GeneBench-Pro Public Release SubsetScience14.08.0−6.0
Eval Connections One-Shot with TrapsGames80.086.0+6.0
WordleBenchGames97.091.0−6.0
VCT-Refusal (CAIS)Safety4.810.7+5.9
BuseyBench SVGMultimodal6.66.1−5.6
GPQA DiamondGeneral QA95.589.9−5.6
Bio Capabilities Index (SecureBio)Indexes162155−5.4
MMMU ProMultimodal84.979.5−5.4
Wiz Cyber Model Arena v3.9Cybersecurity59.164.4+5.3
Excel Modeling BenchmarkFinance & Accounting71.366.2−5.1
Finance Agent v2 - AdjustmentsFinance & Accounting54.749.7−5.0
AA-Briefcase EloAgents11161201+5.0
SciCode 1.0.1 (Artificial Analysis)Code59.855.0−4.9
𝜏³-BankingAgents35.540.2+4.7
Finance Agent v2 - All PassFinance & Accounting47.542.8−4.7
LMCAGeneral QA50.455.0+4.6
GDPval (Mercor 213-task rubric score)Professional Work85.280.7−4.5
MRCR v2 (8-needle, Context Arena) - 128K AUCGeneral QA93.989.4−4.5
Vibe Code Bench v1.1Code70.474.6+4.2
AA-LCR v1.1Search83.079.0−4.0
SkillsBench without SkillsAgents44.948.8+4.0
DeepResearch Bench II (Mercor 130-task set)Search43.447.2+3.8
LMArena Text Arena EloChat & Writing14891468−3.8
GPQA-Bio (SecureBio)Science81.685.3+3.7
VoxelBench Text RatingGames18441894+3.6
HPCT (SecureBio)Science64.768.2+3.5
DTBenchGeneral QA96.893.3−3.5
LMArena Agent Arena Steering BurdenAgents-7.3-0.5+3.4
Creative Writing v3 Rubric ScoreChat & Writing15.215.9+3.4
AIME 2025Math96.7100.0+3.3
Finance Agent v2 - ComparablesFinance & Accounting50.347.0−3.3
LMArena Agent Arena Bash RecoveryAgents-5.40.9+3.2
LAB-Bench LitQA2 (SecureBio)Science54.251.1−3.1
Artificial Analysis Intelligence Index v4.3.2Indexes39.142.1+3.0
Terminal-Bench Science 0.1 (Vals)Science7.110.0+2.9
Artificial Analysis Intelligence Index v4.3Indexes39.442.3+2.8
AHK-EvalCode91.788.9−2.8
LMArena Agent Arena Bash ReliabilityAgents6.20.7−2.7
LiveCodeBench (Vals AI public dataset)Code88.785.9−2.7
APEX-Accounting 1.1Finance & Accounting57.259.8+2.6
Finance Agent v2 - General Qualitative AnalysisFinance & Accounting65.367.8+2.5
APEX-1 (Mercor, mean score)Professional Work67.069.5+2.5
MMLU ProGeneral QA90.688.1−2.5
OTIS Mock AIME 2024-2025Math97.299.7+2.5
RedlineBenchLegal45.347.8+2.5
Continual Learning Exam (Trial 1)Agents25.327.8+2.5
Humanity's Last Exam DiamondGeneral QA31.228.8−2.4
AutomationBench-AA 1.0.6Agents62.059.6−2.4
HackerRank-BenchCode28.526.1−2.4
ACCoRDGeneral QA77.074.7−2.3
Artificial Analysis Progress Index v1Indexes317324+2.3
SAGEMath49.247.0−2.2
LegalBenchLegal87.385.1−2.2
VCT (SecureBio)Science49.347.2−2.1
ObviousBenchGeneral QA100.097.9−2.1
LMArena Agent Arena SteerabilityAgents-0.04.1+2.1
CharXiv Extended (Mercor held-out)Multimodal79.181.1+2.0
Mystery Game PuzzlesGames37.035.0−2.0
Epoch Capabilities Index (ECI)Indexes158161+2.0
MBCT (SecureBio)Science55.056.9+1.9
LMArena Agent Arena Confirmed SuccessAgents-3.5-7.2−1.9
Vals Index v2.1Indexes51.353.1+1.8
WCB (SecureBio)Science58.059.8+1.8
Finance Agent v2 - Earnings AnalysisFinance & Accounting79.177.3−1.8
DAYJOB: HealthcareHealth & Medicine0.01.6+1.6
Conceptual Reasoning IndexIndexes64.665.7+1.5
TaxEval v2Finance & Accounting74.776.2+1.4
CharXiv (descriptive and reasoning combined)Multimodal95.394.1−1.2
Finance Agent v2 - Financial ModelingFinance & Accounting24.223.0−1.1
PCT Political Inconsistency (CAIS)Safety45.146.2−1.1
MedScribeHealth & Medicine83.982.9−1.1
ARC-AGI-1 Semi-PrivateGeneral QA95.596.5+1.0
GDP.pdf (Artificial Analysis, All-pass)Professional Work23.624.6+1.0
DAYJOB: FinanceFinance & Accounting2.81.8−1.0
GDPval-AA v2Professional Work14331415−0.9
Finance Agent v2 - Disclosure AnalysisFinance & Accounting67.766.9−0.8
MMLU-Bio (SecureBio)Science91.490.7−0.7
GDPval-AA v2.1Professional Work14041388−0.7
ARC-AGI-2General QA84.683.9−0.7
MortgageTax (Vals all 1,258 documents)Finance & Accounting66.767.3+0.7
LMArena Agent Arena Net Improvement (Steering Burden)Agents-2.2-1.8+0.7
ProgramBenchCode0.00.5+0.5
LMArena Agent Arena Praise vs ComplaintAgents-2.9-1.9+0.5
Vals Index v2Indexes59.359.6+0.3
BrowseComp Extended (Mercor held-out)Search25.525.3−0.2
LMArena Agent Arena Tool HallucinationAgents0.71.0+0.1
Harvey's Legal Agent Benchmark - Antitrust CompetitionLegal0.00.00.0
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 M&ALegal0.00.00.0
Harvey's Legal Agent Benchmark - Data Privacy CybersecurityLegal0.00.00.0
Harvey's Legal Agent Benchmark - Emerging Companies Venture CapitalLegal0.00.00.0
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 - Funds Asset ManagementLegal0.00.00.0
Harvey's Legal Agent Benchmark - ImmigrationLegal0.00.00.0
Harvey's Legal Agent Benchmark - Intellectual PropertyLegal0.00.00.0
Harvey's Legal Agent Benchmark - International Trade SanctionsLegal0.00.00.0
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 - TaxLegal0.00.00.0
Harvey's Legal Agent Benchmark - Trusts Estates Private ClientLegal0.00.00.0
Harvey's Legal Agent Benchmark - White Collar Defense InvestigationsLegal0.00.00.0

Gemini 3.7 FlashvsGPT-5.6 Terranormalised score marginsData Nov 15, 2026Data: Vals AI, Mercor, Artificial Analysis and 44 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…