Atlas

Compare Models

GPT-5.6 Terra vs GPT-6 Astra

GPT-6 Astra leads by 6.5 capability points (≈95% interval +4.2 to +8.9): a clear lead over GPT-5.6 Terra · P(higher) ≈ 100%

Wins 225 of 272 shared benchmarks (25 lost, 22 tied)

Capability vs. relative cost
GPT-5.6 TerraGPT-6 AstraClosed 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.

GPT-5.6 TerravsGPT-6 Astracapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins25 – 225 · 272 shared, 22 tied

Comparing matching setups only (10 benchmarks excluded) · suites counted once
← GPT-5.6 TerraGPT-6 Astra →

GPT-5.6 TerravsGPT-6 Astrawins by categoryData Nov 15, 2026Data: Vals AI, Arena, Artificial Analysis and 67 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is GPT-6 Astra minus GPT-5.6 Terra.

Concept Synth INDUCTION Challenge100 Round 1General QA25.094.0+69.0
ARC-AGI-3 (Semi-Private)General QA0.862.7+61.9
Concept Synth INDUCTION Challenge64 Round 1General QA37.598.4+60.9
Terminal-Bench Science 0.1 (Vals)Science10.065.7+55.7
ExploitBench - release-reported exploit progressCybersecurity46.2100.0+53.8
Terminal-Bench 4.0 (Artificial Analysis)Code10.159.6+49.5
ChessBench EloGames9251755+49.4
Arena Code WebDev — FullstackCode15651753+49.1
Mystery Game PuzzlesGames35.084.0+49.0
SRE Bench — Fully SolvedCode9.556.9+47.3
EnigmaEvalGeneral QA23.169.7+46.6
Terminal-Bench-Science 0.1Science13.859.5+45.7
Olam Social Poker EloGames15111550+42.4
Agent Red Teaming k=100 (CAIS)Safety54.613.5+41.1
GameBench 2Games53.594.3+40.9
Arena Code WebDev — SimulationsCode15261883+40.3
Odysseys2Agents44.684.6+40.0
Arena Image-to-WebDevCode15391722+39.9
Arena Code WebDev — Reference-Based DesignCode15161807+38.5
NoCoT-Bench NCRI 15.2General QA89.0159+38.5
Arena Code WebDev — Content Creation ToolsCode15031741+37.6
ChartographyMultimodal34.071.0+37.0
ALE-BenchCode21223164+36.9
Arena Code WebDev — HTMLCode15231786+36.8
SnakeBench Average ApplesGames12.719.8+35.9
Arena Code WebDev — ReactCode15111776+35.5
Arena Code WebDev — FrontendCode15211795+35.2
Arena Code WebDev — GamingCode15481861+35.0
Humanity's Last Exam DiamondGeneral QA28.863.8+35.0
TextQuests Game ProgressGames40.574.2+33.7
HardGeoBench Avg DistanceMultimodal1991908+33.4
ZeroBench (Main questions, pass@5)Multimodal19.052.0+33.0
LMArena Agent Arena Net ImprovementAgents0.610.4+32.1
CadQueryEval (Corrected Geometry)Code68.0100.0+32.0
Last Translation Benchmark (LTBv1-eval, blind)Chat & Writing12.244.0+31.8
CyberBench v1.1 OverallCybersecurity72.141.1−31.0
LMArena Agent Arena Net Improvement (Steering Burden)Agents-1.814.5+30.9
Humanity's Sixth SenseMultimodal22.753.6+30.9
Terminal-Bench 4.0Code26.357.1+30.8
Hedge-BenchFinance & Accounting13.844.2+30.4
MindCube TinyMultimodal62.692.7+30.1
SnakeBench Win RateGames50.080.0+30.0
Arena Code WebDev — Brand & MarketingCode15301732+29.3
ZeroBench Main Mean pass@1 (five samples)Multimodal14.243.2+29.0
Gray Swan IPI Attack Success (k=15)Safety37.38.5+28.8
GeneBench-Pro Public Release SubsetScience8.036.0+28.0
Arena Code WebDev — Data & AnalyticsCode14961683+27.9
SpatialViz-BenchMultimodal65.293.1+27.9
SEC-Bench Pro - release-reported cohortCybersecurity57.785.4+27.7
ChessBench CoherenceGames69.797.0+27.3
Chess Puzzles (task version 2.0)Games47.074.0+27.0
FrontierMath Tier 4 (v2) - PrivateMath70.797.6+26.9
RuneBenchGames5.57.3+26.7
AutomationBench 1.0.6 (Zapier)Agents14.841.4+26.6
TxBench Antibody Discovery V0Science31.056.9+25.9
Furniture Assembly BenchmarkMultimodal54.280.0+25.8
ChessBench — Historical Output RepairGames40.766.0+25.3
Arena Code WebDev — Consumer ProductCode15051702+25.0
HPCT Refusal (CAIS)Safety25.050.0+25.0
ProofBench v1.1Math74.099.0+25.0
ZeroBench Main Questions pass^5Multimodal10.035.0+25.0
OpenAI MRCR v2 8-needle - 512K-1MGeneral QA72.596.3+23.8
FACTS Benchmark SuiteGeneral QA52.075.7+23.7
TextQuests Average Cumulative HarmSafety22.229.2−23.7
TxBench Oligo Discovery V0Science33.355.5+22.1
AA-Omniscience IndexIndexes0.143.7+21.8
SnakeBench RatingGames29.837.9+21.7
SWEeper-BenchCode35.056.5+21.5
LisanBench Difficulty-WeightedGames11973113+20.5
Harvey's Legal Agent Benchmark - Corporate M&ALegal0.020.0+20.0
Harvey's Legal Agent Benchmark - Structured Finance SecuritizationLegal20.00.0−20.0
Harvey's Legal Agent Benchmark - Trusts Estates Private ClientLegal0.020.0+20.0
Code MigrationCode47.867.7+19.9
DAYJOB: FinanceFinance & Accounting1.821.5+19.7
Extended NYT Connections 940 (Quadratic v1)Games78.498.1+19.7
FrontierPhysics v0.1Science3.623.2+19.6
HieroglyphBenchMultimodal20.540.0+19.5
Shroombench (mushroom identification, top-1)Multimodal35.054.0+19.0
NoCoT-Bench NCKI (kspine v3)General QA104131+19.0
Blueprint-Bench 2Multimodal0.30.5+18.9
ERQAMultimodal63.682.3+18.7
DeepsecBench — Revalidated V1Code19.137.8+18.7
DeepsecBench — Revalidated V1 (233 issues)Code19.137.6+18.6
FrontierCode 2.0 (Artificial Analysis)Code23.942.4+18.5
Humanity's Last ExamGeneral QA35.953.6+17.7
BenchCAD - geometric overlapMultimodal78.295.9+17.7
SnakeBench Best ApplesGames21.027.0+17.6
APEX-AccountingFinance & Accounting40.357.9+17.6
APEX-SWECode31.148.6+17.5
HackerRank-BenchCode26.143.5+17.4
Roboflow Vision Evals (October 2026, five tasks)Multimodal66.083.4+17.4
GDPval Extended (Mercor held-out rubric score)Professional Work47.264.1+16.9
LMArena Agent Arena Praise vs ComplaintAgents-1.931.7+16.8
Creative Writing v3 EloChat & Writing18562177+16.2
IntPhys 2Multimodal61.377.4+16.1
VCT v2 (SecureBio)Science41.958.0+16.1
GBENCH One-shot Coding GScoreGames63.379.2+15.9
SUPERChemScience64.980.3+15.4
BioSecBench-Function V0Science33.548.6+15.1
BALROGGames53.268.3+15.1
RedlineBenchLegal47.862.8+15.0
WeirdML (v2)Science78.393.3+15.0
BullshitBench v2 LeaderboardSafety54.069.0+15.0
HardGeoBench Country AccuracyMultimodal45.060.0+15.0
HardGeoBench <=100miMultimodal36.351.3+15.0
Revelo Code IndexCode30.645.0+14.4
BioTIER RefusalSafety65.679.9+14.3
MedChemBench - OpenAI internal release scoreScience35.049.3+14.3
ChessBench AccuracyGames71.985.0+14.0
DeepSWE v1.1Code60.274.1+13.9
AA-Briefcase EloAgents13301562+13.5
CritPt v2 (agentic, avg@5)Science69.783.1+13.4
LLM Chess EloGames12931614+13.2
Vibe Code Bench 1–100Code14.827.6+12.8
Roboflow Vision Evals (September 2026 protocol)Multimodal73.886.6+12.8
LisanBench Path LengthGames33037393+12.7
MMMU-Pro Extended (Mercor held-out)Multimodal47.460.0+12.6
MentalHealthBenchSafety46.258.7+12.5
HardGeoBench <=10miMultimodal25.037.5+12.5
BuseyBench SVGMultimodal6.17.3+12.4
Toloka Arena (v1)Agents41.453.8+12.4
IOI (Vals v2)Code87.6100.0+12.4
Bug Hunt BenchCode30.542.9+12.4
LAB-Bench ProtocolQA (SecureBio)Science77.089.3+12.3
Machiavelli Deception (CAIS)Safety80.476.3+12.3
BabaIsBench Stage 0: The IntroGames88.0100.0+12.0
DeepsecBench-AA (Artificial Analysis)Cybersecurity25.136.9+11.9
VCT (SecureBio)Science47.258.5+11.3
Conceptual Reasoning IndexIndexes65.773.9+11.3
LiveBench 2.0Agents75.086.1+11.2
AHK-EvalCode88.9100.0+11.1
ARC-AGI-2General QA83.995.0+11.1
MedXpertQA MM (200-question set)Health & Medicine75.786.8+11.1
SM BenchSafety59.069.9+10.9
LMArena Agent Arena Confirmed SuccessAgents-7.214.0+10.6
OpenAI MRCR v2 8-needle - 256K-512KGeneral QA89.6100.0+10.4
BioSecBench-RefusalSafety34.344.6+10.3
DAYJOB: HealthcareHealth & Medicine1.611.6+10.0
SWE-Marathon v1.1Code32.542.5+10.0
VariantBenchScience38.748.6+9.9
MedXpertQA MM Extended (Mercor held-out)Health & Medicine51.060.8+9.8
Artificial Analysis Intelligence IndexIndexes51.661.2+9.6
LMCAGeneral QA55.064.4+9.4
Senior SWE-BenchCode27.436.8+9.4
Finance Agent v2 - ComparablesFinance & Accounting47.037.8−9.3
Artificial Analysis Progress Index v1Indexes324350+9.0
FrontierCode 1.1 Extended - Pass RateCode61.870.6+8.9
AutomationBench-AA 1.0.6Agents59.668.5+8.8
PCT Political Inconsistency (CAIS)Safety46.237.4+8.8
FrontierCode 1.1 Extended - ScoreCode55.864.5+8.6
Eval Connections One-Shot with TrapsGames86.094.0+8.0
Finance Agent v2 - Earnings AnalysisFinance & Accounting77.369.5−7.8
Harvey HLAB — Criteria Pass RateLegal79.887.5+7.7
FrontierMath Tiers 1-3 (v2) - PrivateMath86.093.7+7.7
LMArena Agent Arena Steering BurdenAgents-0.514.8+7.7
GDP.pdf (Artificial Analysis, All-pass)Professional Work24.632.2+7.6
BioSecBench-RespiratoryVirus-Surveillance V1 (110 evaluations)Science5.813.3+7.6
ScBench-Long V0 VerifiedScience34.942.4+7.6
LAB-Bench CloningScenarios (SecureBio)Science92.5100.0+7.5
Bio Capabilities Index (SecureBio)Indexes155164+7.0
ACCoRDGeneral QA74.781.7+7.0
Vals Index v2Indexes59.666.6+7.0
MBCT (SecureBio)Science56.963.9+7.0
MMMU ProMultimodal79.586.4+6.9
VCT-Refusal (CAIS)Safety10.717.6+6.9
Harvey LAB (Mercor run, undisclosed ~213-task cohort)Legal2.39.2+6.9
Epoch Capabilities Index (ECI)Indexes161169+6.7
Agents' Last Exam (ALE)Agents28.034.2+6.2
Finance Agent v2 - General Quantitative AnalysisFinance & Accounting72.778.8+6.1
GDPval (Mercor 213-task rubric score)Professional Work80.786.7+6.0
HealthBench ProfessionalHealth & Medicine62.468.2+5.8
Wiz Cyber Model Arena v3.9Cybersecurity64.470.2+5.8
RiemannBenchMath66.472.0+5.6
GDPval-AA v2.1Professional Work14611587+5.5
Excel Modeling BenchmarkFinance & Accounting66.271.7+5.5
Finance Agent v2 - PrecedentsFinance & Accounting24.830.3+5.4
SnakeBench Total ApplesGames127198+5.4
Finance Agent v2 - Disclosure AnalysisFinance & Accounting66.961.7−5.3
MASKSafety93.398.5+5.2
MedCodeHealth & Medicine43.448.5+5.1
MedScribeHealth & Medicine82.987.9+5.0
ProgramBenchCode0.55.5+5.0
Finance Agent v2 - Market AnalysisFinance & Accounting60.665.5+4.9
Longform Creative WritingChat & Writing78.082.8+4.8
EpiBenchScience39.944.6+4.7
SWE-Bench Pro V2 — FullCode92.496.9+4.5
Vision Arena OverallMultimodal12661283+4.4
GPQA DiamondGeneral QA91.996.2+4.3
BrowseCompSearch87.591.5+4.0
CharXiv Extended (Mercor held-out)Multimodal81.185.1+4.0
DTBenchGeneral QA93.397.3+4.0
CVDP (517-problem set)Code50.954.9+4.0
AdvancedIFChat & Writing78.182.0+3.9
HealthBench Hard (length-adjusted)Health & Medicine32.736.6+3.9
SciCode (main problems)Science15.819.6+3.8
GDPval-AA v2Professional Work14791555+3.8
Quant-Arena (iterative total score)Finance & Accounting47.651.3+3.7
LAB-Bench LitQA2 (SecureBio)Science51.154.8+3.7
LMArena Agent Arena Bash ReliabilityAgents0.77.8+3.5
Financial Coaching BenchFinance & Accounting71.067.5−3.4
MMLU ProGeneral QA88.191.5+3.4
Medical Long Context Reasoning (MLCR-AA)Health & Medicine31.735.0+3.3
APEX-Agents 1.1Professional Work71.875.1+3.3
SpatialBench-Long V0 VerifiedScience37.934.9−3.0
DeepResearch Bench II (Mercor 130-task set)Search47.250.1+2.9
𝜏³-BankingAgents40.243.1+2.9
WCB (SecureBio)Science59.862.6+2.8
LMArena Agent Arena SteerabilityAgents4.1-1.4−2.8
SWE Atlas - Codebase QnACode45.247.8+2.6
Finance Agent v2 - All PassFinance & Accounting42.840.3−2.5
CRED v0.92Science97.099.5+2.5
ITBench-AAAgents51.048.6−2.4
LMArena Agent Arena Bash RecoveryAgents0.95.5+2.3
SAB RankingForecasting-0.0-0.0−2.1
DistressBenchSafety76.474.3−2.1
GPQA-Bio (SecureBio)Science85.387.4+2.1
ObviousBenchGeneral QA97.9100.0+2.1
ARC-AGI-1 Semi-PrivateGeneral QA96.598.5+2.0
CharXiv (descriptive and reasoning combined)Multimodal94.196.1+2.0
Legal Research BenchLegal41.339.4−1.9
Tax Agent BenchFinance & Accounting65.263.3−1.9
SAB ResearchFinance & Accounting93.495.2+1.8
HealthBench (updated implementation)Health & Medicine58.756.9−1.8
CritPt (avg@5)Science30.031.7+1.7
SpeciEval (Refusal-Aware Retries, Historical)Safety98.196.5−1.5
SpeciEval (Pooled English Runs)Safety98.196.5−1.5
SciCode 1.0.1 (Artificial Analysis)Code55.056.5+1.5
SEA-HELMGeneral QA75.476.9+1.5
Finance Agent v2 - Financial ModelingFinance & Accounting23.021.6−1.4
Finance Agent v2 - AdjustmentsFinance & Accounting49.748.3−1.4
HealthBench (length-adjusted)Health & Medicine57.058.3+1.3
Ruby LLM Benchmarks — Program FixerCode74.075.1+1.2
WMDP-Bio (SecureBio)Science89.790.8+1.1
LMArena Text Arena EloChat & Writing14681474+1.1
AA-LCR v1.1Search79.080.0+1.0
HPCT (SecureBio)Science68.269.0+0.8
PubMedQA (SecureBio)Health & Medicine80.481.2+0.8
SAGEMath47.046.4−0.6
Finance Agent v2 - General Qualitative AnalysisFinance & Accounting67.868.4+0.6
MMLU-Bio (SecureBio)Science90.791.3+0.6
Terminal-Bench 2.1Code88.088.4+0.4
Document ArenaProfessional Work14681467−0.4
SpatialBench V0 VerifiedScience70.470.1−0.3
OTIS Mock AIME 2024-2025Math99.7100.0+0.3
VulcanBench Frontier v4Code89.189.3+0.2
APEX-1 (Mercor, mean score)Professional Work69.569.3−0.2
HealthBench Consensus (updated implementation)Health & Medicine95.295.4+0.2
Creative Writing v3 Rubric ScoreChat & Writing15.915.9+0.2
HealthBench Hard (updated implementation)Health & Medicine34.334.2−0.1
LMArena Agent Arena Tool HallucinationAgents1.01.00.0
AIME 2025Math100.0100.00.0
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 - Banking FinanceLegal0.00.00.0
Harvey's Legal Agent Benchmark - Bankruptcy RestructuringLegal0.00.00.0
Harvey's Legal Agent Benchmark - Capital MarketsLegal0.00.00.0
Harvey's Legal Agent Benchmark - Corporate GovernanceLegal0.00.00.0
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 - Energy Natural ResourcesLegal0.00.00.0
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 - Healthcare Life SciencesLegal0.00.00.0
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 - TaxLegal0.00.00.0
Harvey's Legal Agent Benchmark - White Collar Defense InvestigationsLegal0.00.00.0

GPT-5.6 TerravsGPT-6 Astranormalised score marginsData Nov 15, 2026Data: Vals AI, Arena, Artificial Analysis and 67 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…