Atlas

Compare Models

Claude Opus 5 vs GPT-6 Astra

GPT-6 Astra leads by 3.2 capability points (≈95% interval +0.6 to +5.7): a clear lead over Claude Opus 5 · P(higher) ≈ 99%

Wins 194 of 308 shared benchmarks (86 lost, 28 tied)

Capability vs. relative cost
Claude Opus 5GPT-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.

Claude Opus 5vsGPT-6 Astracapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins86 – 194 · 308 shared, 28 tied

Comparing matching setups only (133 benchmarks excluded) · suites counted once
← Claude Opus 5GPT-6 Astra →

Claude Opus 5vsGPT-6 Astrawins by categoryData Nov 15, 2026Data: Vals AI, benchmarks.bio, Artificial Analysis and 84 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 Claude Opus 5.

Clipping Bench v1.0Games7349514+94.4
Time Horizon Index: KSPAgents18.890.5+71.7
Concept Synth INDUCTION Challenge100 Round 1General QA24.094.0+70.0
ActiveVisionMultimodal10.676.5+65.9
Concept Synth INDUCTION Challenge64 Round 1General QA32.898.4+65.6
Machiavelli Deception (CAIS)Safety96.576.3+61.0
ChessBench EloGames8261755+55.3
HPCT Refusal (CAIS)Safety100.050.0−50.0
Into the Unknown Integrity Node Pass RateScience59.597.8+38.3
Document ArenaProfessional Work15091467−37.0
Terminal-Bench Science 0.1 (Vals)Science27.162.9+35.7
Olam Social Poker EloGames15181550+34.8
Terminal-Bench-Science 0.1Science30.064.6+34.6
GlobeBenchGeneral QA15.949.6+33.6
ChessBench — Historical Output RepairGames32.766.0+33.3
ChessBench CoherenceGames63.897.0+33.2
Chess Puzzles (task version 2.0)Games43.074.0+31.0
EBR-bench v4 (card ban, single agent)Games45.776.2+30.5
Agent Red Teaming k=100 (CAIS)Safety43.313.5+29.8
Odysseys2Agents58.384.6+26.3
EnterpriseBench: CoreCraftProfessional Work68.742.6−26.1
ZeroBench Main Mean pass@1 (five samples)Multimodal17.243.2+26.0
CrossSearch (pass^3)Search34.060.0+26.0
ZeroBench (Main questions, pass@5)Multimodal26.052.0+26.0
EnigmaEvalGeneral QA43.969.7+25.8
MMMU-Pro Extended (Mercor held-out)Multimodal34.360.0+25.7
VIALSMultimodal24.449.9+25.5
HieroglyphBenchMultimodal14.640.0+25.4
Mystery Game PuzzlesGames59.084.0+25.0
LisanBench Path LengthGames154287393−25.0
ARC-AGI-3 (Semi-Private)General QA30.254.8+24.7
Medical Long Context Reasoning (MLCR-AA)Health & Medicine59.435.0−24.4
FrontierMath Tier 4 (v2) - PrivateMath73.297.6+24.4
NoCoT-Bench NCRI 15.2General QA115159+24.1
TaxCalcBench TY25Finance & Accounting20.044.0+24.0
ZeroBench Main Questions pass^5Multimodal11.035.0+24.0
VoxelBench Text RatingGames23002625+23.1
Humanity's Sixth SenseMultimodal30.553.6+23.1
FoodTruck BenchGames7526457761−22.5
TextQuests Average Cumulative HarmSafety22.629.2−22.3
LisanBench Difficulty-WeightedGames50683113−20.9
TextQuests Game ProgressGames53.374.2+20.9
AutoformBenchMath61.982.4+20.5
ComplexConstraintsChat & Writing37.357.7+20.4
BioTIER RefusalSafety100.079.9−20.1
CadQueryEval (Corrected Geometry)Code80.0100.0+20.0
Eval Connections One-Shot with TrapsGames74.094.0+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 - Structured Finance SecuritizationLegal20.00.0−20.0
Harvey's Legal Agent Benchmark - Trusts Estates Private ClientLegal0.020.0+20.0
Blueprint-Bench 2Multimodal0.30.5+19.3
ALE-BenchCode26223164+19.2
Furniture Assembly BenchmarkMultimodal60.880.0+19.2
Humanity's Last Exam DiamondGeneral QA44.863.8+19.0
RedlineBenchLegal44.062.8+18.8
OncologyBench Strict (25 tasks, October 9 rubric)Health & Medicine40.021.3−18.7
SpatialViz-BenchMultimodal75.893.1+17.3
ABC-Bench liquid handling (SecureBio)Science94.076.7−17.3
Vals Science IndexIndexes47.965.1+17.2
LLM Debate BenchmarkChat & Writing17241606−16.9
Into the Unknown Direction Node Pass RateScience66.883.6+16.8
SCI-AGENT-EXAMScience64.280.9+16.7
BioSecBench-Refusal V1 Clean RerunsSafety24.540.8+16.3
FrontierSWE V2 Worst@5Code39.155.1+16.0
SM BenchSafety85.869.9−15.9
Legal Research BenchLegal55.339.4−15.9
MysteryMechanismScience37.453.2+15.8
IOI (Vals v2)Code84.3100.0+15.7
Hedge-BenchFinance & Accounting29.044.2+15.2
BioSecBench-RefusalSafety29.444.6+15.2
APEX-SWECode63.748.6−15.1
MedCodeHealth & Medicine63.648.5−15.1
ChessBench AccuracyGames71.085.0+15.0
AutomationBench 1.0.6 (Zapier)Agents26.941.4+14.5
GeneBench-Pro Public Release SubsetScience22.036.0+14.0
TasteVal (compute multiplier)Science1.31.6+13.9
LMArena Agent Arena Net Improvement (Steering Burden)Agents7.214.5+13.8
LLM Chess EloGames12851614+13.5
Tax Agent BenchFinance & Accounting76.763.3−13.3
TxBench Oligo Discovery V0Science42.255.5+13.3
Terminal-Bench 4.0 (Artificial Analysis)Code46.559.6+13.1
Artificial Analysis Intelligence Index v5.0.1Indexes44.256.9+12.7
Roboflow Vision Evals (October 2026, five tasks)Multimodal70.983.4+12.5
FrontierSWE V2 Best@5Code60.773.0+12.4
IntPhys 2Multimodal65.177.4+12.3
Arena Alignment IndexSafety75.687.8+12.2
LVBenchMultimodal75.487.5+12.1
AutomationBench-AA 1.0.6Agents56.668.5+11.9
VCT-Refusal (CAIS)Safety29.417.6−11.8
CritPt v2 (agentic, avg@5)Science71.483.1+11.7
ERQAMultimodal70.882.3+11.5
SpatialBench-Long V1 Clean RerunsScience28.039.3+11.3
FrontierCode 2.0 (Artificial Analysis)Code31.242.4+11.2
GBENCH One-shot Coding GScoreGames68.179.2+11.1
DeepsecBench-AA (Artificial Analysis)Cybersecurity25.936.9+11.1
DAXBench v2Code86.296.8+10.6
GDP.pdf (Artificial Analysis, All-pass)Professional Work21.632.2+10.6
Code MigrationCode57.567.7+10.3
BuseyBench SVGMultimodal6.37.3+10.2
DAYJOB: FinanceFinance & Accounting11.321.5+10.2
HealthBenchHealth & Medicine67.156.9−10.2
Bio Capabilities Index (SecureBio)Indexes178164−10.0
Shroombench (mushroom identification, top-1)Multimodal44.054.0+10.0
LMArena Agent Arena Praise vs ComplaintAgents12.231.7+9.8
OpenAI internal computer-use safety - adverse outcome rateSafety11.51.8+9.7
BrowseComp (Mercor 130-question subset)Search84.694.2+9.6
HPCT (SecureBio)Science78.669.0−9.6
HardGeoBench Avg DistanceMultimodal1218908+9.6
GameBench 2Games84.894.3+9.5
ClockBench AccuracyMultimodal65.575.0+9.5
PLE (Physical-Engineering's Last Exam)Professional Work38.848.3+9.5
EpiBenchScience35.244.6+9.4
TxBench Oligo Discovery V1 Clean RerunsScience40.049.2+9.2
Snorkel Agentic Coding 2.0Code38.947.6+8.7
SpeciEval (Pooled English Runs)Safety87.996.5+8.6
SpeciEval (Refusal-Aware Retries, Historical)Safety87.996.5+8.6
MaxxBenchCode63.571.9+8.4
MindCube TinyMultimodal84.392.7+8.4
BrokenArXiv 06/2026Math90.799.1+8.3
HiL-Bench (Human-in-Loop Benchmark)Agents57.065.3+8.3
Roboflow Vision Evals (September 2026 protocol)Multimodal78.386.6+8.3
LAB-Bench ProtocolQA (SecureBio)Science81.189.3+8.2
Drone-BenchAgents86.995.1+8.2
NoCoT-Bench NCKI (kspine v3)General QA119131+8.1
MacroscopeBenchCode69.978.0+8.1
FrontierMath Tiers 1-3 (v2) - PrivateMath85.693.7+8.1
Epiq LAER CorporateBenchProfessional Work74.082.0+8.0
GPQA DiamondGeneral QA87.995.8+7.9
GDPval-AA v2Professional Work17081555−7.5
CVDP (517-problem set)Code62.254.9−7.3
VCT (SecureBio)Science51.458.5+7.1
TxBench Antibody Discovery V1 Clean RerunsScience41.948.9+7.1
MedXpertQA MM (200-question set)Health & Medicine79.886.8+7.0
Physical-Engineering's Last Exam (hidden test)Professional Work38.945.8+6.9
Search ArenaSearch12391231−6.6
micro1 PharmaBenchHealth & Medicine9.716.1+6.4
GDPval-AA v2.1Professional Work17291587−6.3
HardGeoBench Country AccuracyMultimodal53.860.0+6.3
WCB (SecureBio)Science68.862.6−6.2
BioSecBench-Function V1 Clean RerunsScience42.548.6+6.2
ScBench-Long V0 VerifiedScience36.442.4+6.1
LMArena Agent Arena SteerabilityAgents10.7-1.4−6.0
DeepResearch Bench II (Mercor 130-task set)Search56.150.1−6.0
BenchCAD - geometric overlapMultimodal89.995.9+6.0
ORCA-bench Hard RCA (Fireworks 324-task subset)Agents44.350.2+5.9
BioSecBench-Structure V1 (58 evaluations)Science15.521.3+5.7
ScBench V1 Clean RerunsScience58.363.9+5.6
BioSecBench-Function V0Science43.048.6+5.6
Learn2Play BenchGames81.876.2−5.6
SUPERChemScience74.780.3+5.6
Artificial Analysis Progress Index v1Indexes334350+5.6
TxBench Antibody Discovery V0Science51.456.9+5.5
SnorkelLegalLegal26.420.9−5.5
BigFinanceBench private holdout (final-answer accuracy)Finance & Accounting49.254.6+5.4
DeepsecBench — Revalidated V1Code32.437.8+5.4
DeepsecBench — Revalidated V1 (233 issues)Code32.337.6+5.3
CharXiv Extended (Mercor held-out)Multimodal79.885.1+5.3
LMArena Agent Arena Steering BurdenAgents4.414.8+5.2
MASKSafety93.498.5+5.1
HardGeoBench <=10miMultimodal32.537.5+5.0
Toloka Arena (v3)Agents62.067.0+5.0
HardGeoBench <=100miMultimodal46.351.3+5.0
SnorkelFinance 2.0Finance & Accounting16.511.6−4.9
BALROGGames63.468.3+4.9
AA-Briefcase EloAgents16451562−4.8
Ruby LLM Benchmarks — Program FixerCode79.875.1−4.7
FORTRESS Average Risk Score (ARS)Safety16.812.1+4.7
GPQA-Bio (SecureBio)Science82.787.4+4.7
ARC-AGI-2General QA90.495.0+4.6
MetagenomicsBench V1Science54.049.7−4.3
MetagenomicsBench V0Science54.049.7−4.3
DistressBenchSafety70.274.3+4.1
LiveBench 2.0Agents82.186.1+4.0
LMArena Agent Arena Bash RecoveryAgents13.65.5−4.0
BullshitBench v2 LeaderboardSafety73.069.0−4.0
RiemannBenchMath68.072.0+4.0
SWE-Clar (implementation)Code63.067.0+4.0
Terminal-Bench 4.0Code53.957.9+4.0
VariantBench V1 Clean RerunsScience50.646.6−4.0
MBABench 2.0Finance & Accounting4.07.9+3.9
Co-Design BenchCode66.169.9+3.8
Extended NYT Connections 940 (Quadratic v1)Games94.398.1+3.8
Epoch Capabilities Index (ECI)Indexes164169+3.8
HealthBench ProfessionalHealth & Medicine73.469.7−3.7
MBCT (SecureBio)Science60.263.9+3.7
ProgramBench (Mercor 100-task subset)Code4.38.0+3.7
TxBench Preclinical Pharmacology V1 Clean RerunsScience63.059.3−3.7
Longform Creative WritingChat & Writing86.382.8−3.5
SWE-Clar (clarification)Code84.087.5+3.5
SpatialBench V0 VerifiedScience66.770.1+3.5
Agents' Last Exam Overall ScoreAgents55.959.3+3.4
APEX-AccountingFinance & Accounting54.557.9+3.4
LAB-Bench LitQA2 (SecureBio)Science51.454.8+3.4
SAB RankingForecasting-0.1-0.0+3.4
AA-Omniscience IndexIndexes37.143.7+3.3
Harvey LAB: In-House Contracting (held-out)Legal2.05.3+3.3
MedXpertQA MM Extended (Mercor held-out)Health & Medicine64.160.8−3.3
LLM Creative Story WritingChat & Writing3.73.4−3.2
SWE-TogetherCode63.860.6−3.2
DAYJOB: HealthcareHealth & Medicine8.411.6+3.2
EpiBench V1 Clean RerunsScience25.228.3+3.1
MedScribeHealth & Medicine91.087.9−3.1
SAGEMath49.446.4−3.1
SpatialBench-Long V0 VerifiedScience31.834.9+3.0
BioSecBench-Surveillance V0 (102 tasks)Science51.548.5−3.0
SimpleBenchGeneral QA80.683.6+3.0
WitnessBench v1.0Games30.627.6−3.0
BioSecBench-Surveillance V1 Clean RerunsScience51.548.5−3.0
FrontierMath ErdősMath0.02.9+2.9
SnorkelUnderwrite 2.0Finance & Accounting18.321.2+2.9
MA-BenchProfessional Work56.253.4−2.8
Humanity's Last ExamGeneral QA56.453.6−2.8
LMArena Text Arena EloChat & Writing14901474−2.7
LMArena Agent Arena Net ImprovementAgents9.510.4+2.7
FrontierSWE V2 (Mercor rerun, mean@3)Code27.329.9+2.6
CritPt (avg@5)Science29.131.7+2.6
AA-AnalystAgent (pass^5)Professional Work53.851.3−2.5
MMMU ProMultimodal84.086.4+2.4
OSWorld 2 (Aug 8) offline - reported partial reward, step limit unspecifiedAgents70.272.6+2.4
SciCode (main problems)Science17.319.6+2.3
Terminal-Bench 2.1Code87.689.9+2.2
Harvey LAB held-out set (Fireworks run)Legal7.45.1−2.2
ACCoRDGeneral QA83.881.7−2.1
Creative Writing v3 EloChat & Writing21352177+2.1
Artificial Analysis Intelligence Index v4.3Indexes50.752.8+2.1
AutoResearchExam (hidden-test AUARC)Agents0.60.6+2.1
Senior SWE-BenchCode34.736.8+2.1
AA-LCR v1.1Search82.080.0−2.0
VulnPR-100Code48.046.0−2.0
LMArena Agent Arena Confirmed SuccessAgents10.214.0+1.9
SciUniverse Level 1Science30.532.5+1.9
APEX-Agents 1.1Professional Work77.075.1−1.9
Artificial Analysis Intelligence Index v4.3.2Indexes50.852.7+1.9
Excel Modeling BenchmarkFinance & Accounting73.671.7−1.9
𝜏³-BankingAgents44.743.1−1.6
Creative Writing v3 Rubric ScoreChat & Writing16.315.9−1.6
Snorkel SWE-bench CLICode14.012.4−1.6
AdvancedIFChat & Writing80.582.0+1.5
CharXiv (descriptive and reasoning combined)Multimodal94.696.1+1.5
matplotgymCode56.558.0+1.5
WeirdML (v2)Science91.893.3+1.5
Bench to the Future 3 — September Panel — PooledForecasting0.10.1−1.4
SnorkelRevOpsProfessional Work12.210.9−1.3
PostTrainBench Average (v1.1)Indexes45.644.3−1.3
Snorkel Digital Work IndexIndexes21.823.0+1.3
BioSecBench-RespiratoryVirus-Surveillance V1 (110 evaluations)Science14.613.3−1.2
LABBench2Science84.285.4+1.2
GDPval Extended (Mercor held-out rubric score)Professional Work62.964.1+1.2
Bedside BenchHealth & Medicine94.693.4−1.2
OTIS Mock AIME 2024-2025Math98.9100.0+1.1
GDPval (Mercor 213-task rubric score)Professional Work85.686.7+1.1
Conceptual Reasoning IndexIndexes74.773.9−1.1
SAB ResearchFinance & Accounting96.295.2−1.0
ARC-AGI-1 Semi-PrivateGeneral QA97.598.5+1.0
ATLAS-Finance v1.0Finance & Accounting12.311.3−1.0
ATLAS-Finance v1.1Finance & Accounting10.79.7−1.0
Wiz Cyber Model Arena v3.9Cybersecurity71.270.2−1.0
LMArena Agent Arena Bash ReliabilityAgents9.87.8−1.0
DeepSWE v1.1Code73.274.1+1.0
HackerRank-BenchCode42.643.5+0.9
ForecastBench Preliminary Dataset Brier IndexForecasting62.062.9+0.9
Toloka Arena (v1)Agents52.953.8+0.9
Vibe Code Bench 1–100Code28.527.6−0.9
SpatialBench V1 Clean RerunsScience68.467.5−0.9
VariantBenchScience49.448.6−0.8
APEX-1 (Mercor, mean score)Professional Work68.669.3+0.7
PCT Political Inconsistency (CAIS)Safety36.737.4−0.7
SnorkelManufacturingProfessional Work9.410.0+0.6
DTBenchGeneral QA97.997.3−0.5
Vision Arena OverallMultimodal12851283−0.5
MMLU ProGeneral QA92.091.5−0.5
Snorkel WorkplaceAgentsProfessional Work14.914.4−0.5
CheatBenchSafety45.345.0+0.4
ForecastBench Tournament Overall Brier IndexForecasting66.666.9+0.3
Harvey HLAB — Criteria Pass RateLegal87.787.5−0.2
MMLU-Bio (SecureBio)Science91.191.3+0.2
SciCode 1.0.1 (Artificial Analysis)Code56.456.5+0.1
LMCAGeneral QA64.564.4−0.1
LMArena Agent Arena Tool HallucinationAgents0.91.00.0
AIME 2025Math100.0100.00.0
BabaIsBench Stage 0: The IntroGames100.0100.00.0
BioMysteryBench (Vals Terminus 2)Science79.379.30.0
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 - Corporate M&ALegal20.020.00.0
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 - 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
Into the UnknownScience0.00.00.0
LAB-Bench CloningScenarios (SecureBio)Science100.0100.00.0
ObviousBenchGeneral QA100.0100.00.0
ProofBench v1.1Math99.099.00.0
ScBench-Long V1 Clean RerunsScience27.327.30.0

Claude Opus 5vsGPT-6 Astranormalised score marginsData Nov 15, 2026Data: Vals AI, benchmarks.bio, Artificial Analysis and 84 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…