Atlas

Compare Models

Gemini 3.8 Flash vs GPT-5.6 Terra

GPT-5.6 Terra leads by 0.0 capability points (≈95% interval −2.1 to +2.1): statistically tied · P(higher) ≈ 50%

Gemini 3.8 Flash wins 125 of 221 shared benchmarks (79 lost, 17 tied)

Capability vs. relative cost
Gemini 3.8 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 3.8 FlashvsGPT-5.6 Terracapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins125 – 79 · 221 shared, 17 tied

Comparing matching setups only (74 benchmarks excluded) · suites counted once
← Gemini 3.8 FlashGPT-5.6 Terra →

Gemini 3.8 FlashvsGPT-5.6 Terrawins by categoryData Nov 15, 2026Data: Vals AI, Mercor, Artificial Analysis and 48 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 3.8 Flash.

FrontierMath Tier 4 (v2) - PrivateMath22.070.7+48.8
Harvey's Legal Agent Benchmark - Emerging Companies Venture CapitalLegal40.00.0−40.0
ChessBench EloGames1527925−35.8
Olam Social Poker EloGames14791511+34.8
MASKSafety58.693.3+34.7
HieroglyphBenchMultimodal54.620.5−34.1
HLE Extended (Mercor held-out)General QA36.870.7+33.9
Gray Swan IPI Attack Success (k=15)Safety5.537.3−31.8
IOI (Vals v2)Code56.987.6+30.7
Shroombench (mushroom identification, top-1)Multimodal65.035.0−30.0
MindCube TinyMultimodal92.262.6−29.6
CyberBench v1.1 OverallCybersecurity43.872.1+28.3
CadQueryEval (Corrected Geometry)Code96.068.0−28.0
Last Translation Benchmark (LTBv1-eval, blind)Chat & Writing39.812.2−27.7
ProofBench v1.1Math48.074.0+26.0
ChessBench CoherenceGames94.669.7−24.9
Agent Red Teaming k=100 (CAIS)Safety29.854.6−24.8
SWE-bench Verified Extended (Mercor held-out)Code54.078.7+24.7
Concept Synth INDUCTION Challenge64 Round 1General QA14.137.5+23.4
Furniture Assembly BenchmarkMultimodal31.754.2+22.5
SM BenchSafety80.459.0−21.4
NoCoT-Bench NCRI 15.2General QA12689.0−20.2
DeepSWE v1.1Code73.853.8−20.1
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 - Energy Natural ResourcesLegal20.00.0−20.0
Harvey's Legal Agent Benchmark - Healthcare Life SciencesLegal20.00.0−20.0
Harvey's Legal Agent Benchmark - Trusts Estates Private ClientLegal20.00.0−20.0
Extended NYT Connections 940 (Quadratic v1)Games97.478.4−19.0
Humanity's Sixth SenseMultimodal41.622.7−18.9
AA-LCR Extended (Mercor held-out)Search55.674.4+18.8
Machiavelli Deception (CAIS)Safety86.580.4+18.6
BioTIER RefusalSafety83.965.6−18.3
Terminal-Bench 4.0 (Artificial Analysis)Code19.71.5−18.2
Finance Agent v2 - Market AnalysisFinance & Accounting78.560.6−17.9
FrontierMath Tiers 1-3 (v2) - PrivateMath68.486.0+17.5
FACTS Benchmark SuiteGeneral QA69.452.0−17.4
BrowseComp Extended (Mercor held-out)Search41.525.3−16.2
ERQAMultimodal79.663.6−16.0
Concept Synth INDUCTION Challenge100 Round 1General QA9.025.0+16.0
SWE-bench VerifiedCode80.095.4+15.4
RiemannBenchMath51.266.4+15.2
AutomationBench 1.0.6 (Zapier)Agents29.714.8−14.9
AA-Omniscience IndexIndexes29.60.1−14.8
EnigmaEvalGeneral QA37.223.1−14.1
Roboflow Vision Evals (October 2026, five tasks)Multimodal79.866.0−13.8
Vending-Bench 2Games50947343+13.6
TextQuests Game ProgressGames54.140.5−13.6
GDPval Extended (Mercor held-out rubric score)Professional Work60.447.2−13.2
SpatialViz-BenchMultimodal78.165.2−12.9
Document ArenaProfessional Work14821468−12.8
BabaIsBench Stage 0: The IntroGames100.088.0−12.0
HPCT Refusal (CAIS)Safety13.025.0+12.0
Mystery Game PuzzlesGames47.035.0−12.0
Terminal-Bench 2.1Code87.675.7−12.0
SAGEMath35.147.0+11.9
NoCoT-Bench NCKI (kspine v3)General QA121104−11.7
CritPt (avg@5)Science18.330.0+11.7
APEX-AccountingFinance & Accounting51.740.3−11.4
Roboflow Vision Evals (September 2026 protocol)Multimodal85.173.8−11.3
Finance Agent v2 - PrecedentsFinance & Accounting36.124.8−11.3
Code MigrationCode36.547.8+11.3
MMMU-Pro Extended (Mercor held-out)Multimodal36.447.4+11.0
Chess Puzzles (task version 2.0)Games58.047.0−11.0
GameBench 2Games64.053.5−10.5
DeepsecBench-AA (Artificial Analysis)Cybersecurity14.525.1+10.5
Harvey HLAB — Criteria Pass RateLegal90.279.8−10.4
LLM Chess EloGames15461293−10.4
Humanity's Last ExamGeneral QA46.235.9−10.3
LMArena Agent Arena Net ImprovementAgents3.80.6−10.3
Design Arena Overall Code CategoriesCode12991238−10.1
Medical Long Context Reasoning (MLCR-AA)Health & Medicine21.731.7+10.0
ARC-AGI-3 (Semi-Private)General QA10.40.5−9.9
ChessBench AccuracyGames81.171.9−9.9
MedXpertQA MM Extended (Mercor held-out)Health & Medicine60.851.0−9.8
Finance Agent v2 - Financial ModelingFinance & Accounting32.623.0−9.6
Harvey LAB (Mercor run, undisclosed ~213-task cohort)Legal11.42.3−9.1
Hedge-BenchFinance & Accounting4.713.8+9.1
MedXpertQA MM (200-question set)Health & Medicine84.775.7−9.0
SWE-bench MultilingualCode85.894.4+8.6
SpeciEval (Refusal-Aware Retries, Historical)Safety89.998.1+8.2
SpeciEval (Pooled English Runs)Safety89.998.1+8.2
Finance Agent v2 - General Quantitative AnalysisFinance & Accounting80.972.7−8.2
Terminal-Bench 2.1 Extended (Mercor held-out)Code44.436.4−8.0
MBCT (SecureBio)Science48.956.9+8.0
LMArena Agent Arena Net Improvement (Steering Burden)Agents2.3-1.8−7.8
Blueprint-Bench 2Multimodal0.40.3−7.8
SkillsBench without SkillsAgents41.048.8+7.8
SUPERChemScience72.664.9−7.7
LiveBench 2.0Agents67.375.0+7.6
LAB-Bench CloningScenarios (SecureBio)Science100.092.5−7.5
TextQuests Average Cumulative HarmSafety24.422.2+7.3
LAB-Bench ProtocolQA (SecureBio)Science84.077.0−7.0
WordleBenchGames98.091.0−7.0
Finance Agent v2 - All PassFinance & Accounting49.742.8−6.9
ChartographyMultimodal40.934.0−6.9
Surface Evolver BenchAgents76.983.8+6.9
Vision Arena OverallMultimodal12931266−6.8
VoxelBench Text RatingGames17981894+6.8
Continual Learning Exam (Trial 1)Agents21.227.8+6.7
WeirdML (v2)Science84.878.3−6.5
RedlineBenchLegal41.347.8+6.5
Excel Modeling BenchmarkFinance & Accounting72.266.2−6.0
HPCT (SecureBio)Science62.468.2+5.8
DistressBenchSafety70.776.4+5.7
Creative Writing v3 EloChat & Writing17441856+5.7
𝜏³-BankingAgents45.840.2−5.6
BuseyBench SVGMultimodal6.66.1−5.5
APEX-SWECode36.431.1−5.3
ARC-AGI-2General QA89.283.9−5.3
LMArena Text Arena EloChat & Writing14971468−5.2
FrontierCode 2.0 (Artificial Analysis)Code18.723.9+5.1
Finance Agent v2 - AdjustmentsFinance & Accounting54.849.7−5.1
GPQA DiamondGeneral QA94.989.9−5.0
MMMU ProMultimodal84.579.5−5.0
AA-LCR v1.1Search84.079.0−5.0
Finance Agent v2 - ComparablesFinance & Accounting52.047.0−5.0
MedCodeHealth & Medicine48.143.4−4.7
Financial Coaching BenchFinance & Accounting66.371.0+4.7
GBENCH One-shot Coding GScoreGames58.663.3+4.7
SWE Atlas - Codebase QnACode40.645.2+4.6
CritPt v2 (agentic, avg@5)Science65.169.7+4.6
LMArena Agent Arena Bash ReliabilityAgents9.90.7−4.6
Quant-Arena (iterative total score)Finance & Accounting43.047.6+4.5
VCT-Refusal (CAIS)Safety6.410.7+4.3
LMArena Agent Arena Confirmed SuccessAgents1.3-7.2−4.3
Finance Agent v2 - Earnings AnalysisFinance & Accounting81.477.3−4.1
Vibe Code Bench v1.1Code78.774.6−4.1
LiveCodeBench (Vals AI public dataset)Code89.585.9−3.6
ACCoRDGeneral QA78.274.7−3.5
SpatialBench V0 VerifiedScience67.070.4+3.5
LMArena Agent Arena Praise vs ComplaintAgents4.9-1.9−3.4
Terminal-Bench-Science 0.1Science10.513.8+3.3
VCT v2 (SecureBio)Science45.241.9−3.3
AIME 2025Math96.7100.0+3.3
RCM-Bench-1Health & Medicine90.287.0−3.3
Terminal-Bench 4.0Code19.722.7+3.0
GeneBench-Pro Public Release SubsetScience11.08.0−3.0
Public Benefits Bench v1.1Professional Work65.362.4−2.9
GDPval (Mercor 213-task rubric score)Professional Work83.580.7−2.8
BrowseComp (Mercor 130-question subset)Search83.185.8+2.7
WCB (SecureBio)Science57.159.8+2.7
Vals Index v2Indexes62.359.6−2.7
MMLU ProGeneral QA90.788.1−2.6
SWE-Bench Pro V2 — FullCode94.992.4−2.5
Legal Research BenchLegal38.941.3+2.4
DTBenchGeneral QA95.793.3−2.4
GDPval-AA v2Professional Work14641415−2.4
GDPval-AA v2.1Professional Work14421388−2.4
CharXiv Extended (Mercor held-out)Multimodal78.981.1+2.2
ObviousBenchGeneral QA100.097.9−2.1
LMCAGeneral QA52.955.0+2.0
ARC-AGI-1 Semi-PrivateGeneral QA98.596.5−2.0
Eval Connections One-Shot with TrapsGames88.086.0−2.0
MortgageTax (Vals all 1,258 documents)Finance & Accounting65.367.3+2.0
CRM-Agent-BenchAgents83.681.6−2.0
LegalBenchLegal87.085.1−1.9
SEA-HELMGeneral QA77.375.4−1.9
Vals Index v2.1Indexes54.953.1−1.8
GDP.pdf (Artificial Analysis, All-pass)Professional Work22.824.6+1.8
Epoch Capabilities Index (ECI)Indexes159161+1.8
TaxEval v2Finance & Accounting74.476.2+1.7
MedScribeHealth & Medicine84.582.9−1.6
SciCode 1.0.1 (Artificial Analysis)Code56.655.0−1.6
SciCode (main problems)Science14.215.8+1.6
Tax Agent BenchFinance & Accounting66.865.2−1.6
Finance Agent v2 - Disclosure AnalysisFinance & Accounting68.566.9−1.5
BioSecBench-RespiratoryVirus-Surveillance V1 (110 evaluations)Science7.35.8−1.5
ITBench-AAAgents52.551.0−1.5
Creative Writing v3 Rubric ScoreChat & Writing15.615.9+1.5
Artificial Analysis Intelligence Index v5.0.1Indexes30.832.3+1.5
Terminal-Bench Science 0.1 (Vals)Science8.610.0+1.4
CharXiv (descriptive and reasoning combined)Multimodal95.494.1−1.3
Artificial Analysis Progress Index v1Indexes320324+1.3
AutomationBench-AA 1.0.6Agents60.959.6−1.2
IntPhys 2Multimodal62.561.3−1.2
SAGE v1.1Math88.287.0−1.2
Longform Creative WritingChat & Writing76.878.0+1.2
Artificial Analysis Intelligence Index v4.3.2Indexes40.942.1+1.2
APEX-Agents 1.1Professional Work72.971.8−1.1
Artificial Analysis Intelligence Index v4.3Indexes41.242.3+1.1
DAYJOB: FinanceFinance & Accounting2.81.8−1.0
ALE-BenchCode21822154−1.0
SkillsBench with SkillsAgents58.058.9+0.9
Bio Capabilities Index (SecureBio)Indexes153155+0.9
OTIS Mock AIME 2024-2025Math98.999.7+0.8
DAYJOB: HealthcareHealth & Medicine0.81.6+0.8
LMArena Agent Arena Bash RecoveryAgents-0.70.9+0.8
APEX-1 (Mercor, mean score)Professional Work68.869.5+0.7
ProgramBenchCode1.00.5−0.5
Political Science LLM BenchmarkProfessional Work68.768.3−0.4
LAB-Bench LitQA2 (SecureBio)Science51.551.1−0.4
LMArena Agent Arena Steering BurdenAgents-1.3-0.5+0.4
VCT (SecureBio)Science46.947.2+0.3
Humanity's Last Exam DiamondGeneral QA29.128.8−0.3
WMDP-Bio (SecureBio)Science89.489.7+0.3
Finance Agent v2 - General Qualitative AnalysisFinance & Accounting68.167.8−0.3
LMArena Agent Arena Tool HallucinationAgents0.41.0+0.3
CVDP (517-problem set)Code50.750.9+0.2
PCT Political Inconsistency (CAIS)Safety46.346.2+0.1
LMArena Agent Arena SteerabilityAgents4.34.1−0.1
AA-Briefcase EloAgents120212010.0
Conceptual Reasoning IndexIndexes65.765.70.0
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 - 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 - InsuranceLegal0.00.00.0
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 - Structured Finance SecuritizationLegal20.020.00.0
Harvey's Legal Agent Benchmark - TaxLegal0.00.00.0
Harvey's Legal Agent Benchmark - White Collar Defense InvestigationsLegal0.00.00.0

Gemini 3.8 FlashvsGPT-5.6 Terranormalised score marginsData Nov 15, 2026Data: Vals AI, Mercor, Artificial Analysis and 48 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…