Atlas

Compare Models

Claude Opus 5.5 vs Claude Sonnet 5.5

Claude Opus 5.5 leads by 1.1 capability points (≈95% interval −1.5 to +3.8): statistically tied · P(higher) ≈ 80%

Wins 180 of 283 shared benchmarks (72 lost, 31 tied)

Capability vs. relative cost
Claude Opus 5.5Claude Sonnet 5.5Closed 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 5.5vsClaude Sonnet 5.5capability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins180 – 72 · 283 shared, 31 tied

Comparing matching setups only (20 benchmarks excluded) · suites counted once
← Claude Opus 5.5Claude Sonnet 5.5 →

Claude Opus 5.5vsClaude Sonnet 5.5wins by categoryData Nov 15, 2026Data: Vals AI, Mercor, Artificial Analysis and 86 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Claude Sonnet 5.5 minus Claude Opus 5.5.

TxBench Oligo Discovery V1 (113 evaluations)Science1.551.9+50.5
TxBench Antibody Discovery V1 (100 evaluations)Science8.057.3+49.3
YC-BenchGames28737251741315−39.5
SPU (Science Prediction Universe)Science-0.30.3+33.8
LisanBench Difficulty-WeightedGames59632983−31.9
LisanBench Path LengthGames2402913787−31.8
TxBench Preclinical Pharmacology V1 Clean RerunsScience30.762.3+31.7
ViBench 2.0Code48.519.1−29.4
ChessBench EloGames1349857−29.3
ChessBenchGames65.339.3−26.0
SimpleQA Verified (Epoch anti-abstention)General QA72.246.5−25.7
TaxCalcBench TY25Finance & Accounting42.018.0−24.0
TreasuryQA BenchProfessional Work64.841.9−22.9
BioSecBench-Function V1 Clean RerunsScience73.651.7−21.9
Concept Synth INDUCTION Challenge64 Round 1General QA71.950.0−21.9
Olam Social Poker EloGames15061526+21.7
Concept Synth INDUCTION Challenge100 Round 1General QA65.044.0−21.0
Harvey's Legal Agent Benchmark - Data Privacy CybersecurityLegal0.020.0+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 ClientLegal0.020.0+20.0
ScBench-Long V1 Clean RerunsScience13.631.8+18.2
RuneBenchGames6.75.5−17.5
AA-PersonalityBench Expressiveness indexChat & Writing90.573.3−17.3
SpatialBench-Long V1 Clean RerunsScience19.636.5+16.9
MetagenomicsBench V1Science43.760.0+16.3
MetagenomicsBench V0Science43.760.0+16.3
LLM Chess EloGames15751195−15.6
MedXpertQA MM Extended (Mercor held-out)Health & Medicine71.255.6−15.6
ChessBench CoherenceGames80.164.6−15.5
LAB-Bench LitQA2 (SecureBio)Science71.155.8−15.3
FrontierMath Tier 4 (v2) - PrivateMath95.080.5−14.5
SimpleBenchGeneral QA88.475.9−12.5
Clipping Bench v1.0Games2018886−12.2
WCB (SecureBio)Science54.566.5+12.0
Agent Initiative BenchmarkAgents7.86.7−11.9
Blueprint-Bench 2Multimodal0.50.4−11.9
APEX-1 (Mercor, mean score)Professional Work65.853.9−11.9
SWE-bench Multimodal (Vals v1, 480 Tasks)Code70.258.3−11.9
SAFE-Teen 106 worst branch: Relationships (no prompt)Safety25.337.0−11.7
AA-PersonalityBench Conviction indexChat & Writing69.558.1−11.5
AA-PersonalityBench Warmth indexChat & Writing84.873.3−11.5
WeirdML v3Science31.219.9−11.3
Toloka Arena (v3)Agents75.064.0−11.0
DeepsecBench — Revalidated V1 (233 issues)Code26.637.5+10.9
KernelBench CUDA Normalized (RTX PRO 6000)Code76.387.0+10.8
ALE-BenchCode27482448−10.6
Cua Driver Bench v2 (macOS, foreground)Agents88.078.0−10.0
Eval Connections One-Shot with TrapsGames86.076.0−10.0
HoneyBench v0.2 (hack rate)Safety31.641.6−10.0
MBCT (SecureBio)Science68.758.8−9.9
SWE Atlas v1.1 - OverallCode77.467.5−9.9
AA-PersonalityBench Imagination indexChat & Writing94.384.8−9.6
AA-PersonalityBench Transparency indexChat & Writing90.5100+9.6
DAXBench v2Code95.786.2−9.6
MaxxBenchCode76.466.9−9.6
Terminal-Bench-Science 0.1Science61.952.4−9.5
DAYJOB: HealthcareHealth & Medicine24.715.2−9.5
ARC-AGI-2General QA93.383.9−9.4
KernelBench Mega Normalized (RTX PRO 6000)Code100.090.6−9.4
Bug Hunt BenchCode39.748.9+9.2
BioSecBench-Structure V1 (58 evaluations)Science31.021.8−9.2
GeneBench-Pro Public Release SubsetScience30.039.0+9.0
VCT (SecureBio)Science61.452.5−8.9
EpiBench V1 Clean RerunsScience39.030.2−8.8
HardGeoBench <=100miMultimodal62.553.8−8.7
SWE-bench Verified Extended (Mercor held-out)Code66.057.3−8.7
FrontierML ($200 wallet)Agents58.850.4−8.4
Furniture Assembly BenchmarkMultimodal83.375.0−8.3
Medical Long Context Reasoning (MLCR-AA)Health & Medicine66.775.0+8.3
VCT v2 (SecureBio)Science60.552.2−8.3
MineBench — Bradley-Terry RatingGames21341998−8.2
BioSecBench-RespiratoryVirus-Surveillance V1 (110 evaluations)Science25.417.3−8.2
Ramp Accounting BenchFinance & Accounting50.542.5−8.1
Chess Puzzles (task version 2.0)Games67.059.0−8.0
Cua Driver Bench v2 (Linux, foreground)Agents76.084.0+8.0
Extended NYT Connections 940 (Quadratic v1)Games88.580.5−8.0
Decision-Bench v1.1Professional Work59.151.2−8.0
ExploitGymCybersecurity34.526.6−7.9
AA-PersonalityBench Slop indexChat & Writing20.027.7+7.7
Code2Video-AA EloCode15361443−7.5
AA-Omniscience IndexIndexes46.432.3−7.1
PLE (Physical-Engineering's Last Exam)Professional Work50.643.6−7.0
LMArena Agent Arena Net Improvement (Steering Burden)Agents19.516.0−6.7
BigFinanceBench private holdout (final-answer accuracy)Finance & Accounting48.241.5−6.7
MatFlowBench Pass@1Science56.750.0−6.7
PubMedQA (SecureBio)Health & Medicine87.981.3−6.6
GBENCH One-shot Coding GScoreGames73.967.4−6.5
SWE-TogetherCode68.862.4−6.4
HardGeoBench Country AccuracyMultimodal68.862.5−6.2
o11y-bench 1.1 Pass^3Agents84.490.6+6.3
CritPt v2 (agentic, avg@5)Science81.775.7−6.0
Mystery Game PuzzlesGames71.065.0−6.0
SAGEMath45.851.8+5.9
SciCode 1.0.1 (Artificial Analysis)Code66.961.0−5.9
LMArena Text Arena EloChat & Writing15061472−5.9
Harvey HLAB — Criteria Pass RateLegal75.870.0−5.8
Design Arena Overall Code CategoriesCode13841349−5.8
AA-PersonalityBench Organization indexChat & Writing79.184.8+5.8
AA-PersonalityBench Informativeness indexChat & Writing90.584.8−5.8
LAB-Bench ProtocolQA (SecureBio)Science94.088.3−5.7
Real-SWE 1.1Code56.851.2−5.6
LMArena Code Arena WebDevCode18141773−5.5
FrontierCode 2.0 - Pass RateCode51.846.4−5.5
OncologyBench Strict (25 tasks, October 9 rubric)Health & Medicine37.332.0−5.3
FrontierCode 2.0 - ScoreCode47.342.2−5.1
HealthBench (length-adjusted, Artificial Analysis)Health & Medicine57.062.1+5.1
AA-LCR v1.1Search84.779.7−5.0
MazeBench ASCII with PythonGames13.08.0−5.0
MedXpertQA MM (200-question set)Health & Medicine85.780.7−5.0
SWE-Marathon v1.1Code46.351.3+5.0
ScBench V1 Clean RerunsScience55.460.3+5.0
SAFE-Teen 106 worst branch: Relationships (under-18 prompt)Safety2.47.3−4.9
LMArena Agent Arena Praise vs ComplaintAgents40.330.5−4.9
GDPval Extended (Mercor held-out rubric score)Professional Work75.870.9−4.9
VulcanBench Frontier v4Code86.481.6−4.7
ChartographyMultimodal66.361.6−4.7
SAFE-Teen 106: Relationships (no prompt)Safety12.216.9−4.7
HANDBOOK.md Agents v1.1Professional Work52.347.7−4.6
Nimble Web ContextSearch71.867.2−4.6
SWE-bench MultilingualCode95.590.9−4.6
Cresta-BenchAgents82.678.0−4.5
SpireClimbGames5.81.3−4.5
Roboflow Vision Evals (October 2026, five tasks)Multimodal82.077.6−4.4
LiveBench 2.0Agents86.582.5−4.1
Cua Driver Bench v2 (Windows, foreground)Agents74.070.0−4.0
WellnessBenchSafety53.057.0+4.0
VibeVideoBench (overall quality, no video = 0)Code4.13.9−3.9
GDPxProfessional Work12.08.2−3.8
Terminal-Bench 2.1Code83.587.3+3.8
ChessBench AccuracyGames79.375.8−3.8
HardGeoBench <=10miMultimodal43.840.0−3.7
SpeciEval (Pooled English Runs)Safety87.291.0+3.8
SpeciEval (Refusal-Aware Retries, Historical)Safety87.291.0+3.8
WorkflowArenaAgents37.033.2−3.7
Ruby LLM Benchmarks — Program FixerCode77.573.8−3.7
SM BenchSafety81.885.4+3.7
EnterpriseBench: CoreCraftProfessional Work74.470.8−3.6
WMDP-Bio (SecureBio)Science96.192.6−3.5
CVDP (517-problem set)Code64.160.6−3.5
Short Form Bench v1.0Games23892355−3.4
Quant-Arena (iterative total score)Finance & Accounting44.340.9−3.4
GameBench 2Games94.891.4−3.3
GPQA-Bio (SecureBio)Science90.987.6−3.3
OSWorld 2.1 offline - reported partial reward, step limit unspecifiedAgents87.283.9−3.3
SAFE-Teen 106 worst branch: Urgent Safety (no prompt)Safety26.523.2+3.3
SAGE v1.1Math91.388.0−3.3
Hermes BenchAgents76.773.5−3.3
BuseyBench SVGMultimodal6.86.5−3.2
RiemannBench (November 2026 rescoring)Math80.076.8−3.2
Immigration BenchLegal72.169.0−3.2
MedCodeHealth & Medicine49.852.9+3.1
Bio Capabilities Index (SecureBio)Indexes171175+3.0
SAFE-Teen 106 worst branch: Health Advice (under-18 prompt)Safety27.730.8−3.0
Terminal-Bench 4.0Code64.861.8−3.0
GDP.pdf (Artificial Analysis, All-pass)Professional Work28.825.8−3.0
BullshitBench v2 LeaderboardSafety63.060.0−3.0
Physical-Engineering's Last Exam (hidden test)Professional Work47.744.9−2.8
ComplexConstraints (November 2026 rescoring)Chat & Writing64.467.2+2.8
Qodo AI Code Review Benchmark (125 PRs)Code56.854.0−2.8
ACCoRDGeneral QA88.485.7−2.8
Android Bench 2.0Code37.340.0+2.7
FrontierCS Algorithmic v1.1Code56.553.9−2.6
OpenWorldEngineer (Mean@8)Code41.639.0−2.6
Humanity's Last Exam (Text-Only)General QA57.555.0−2.5
Terminal-Bench 4.0 (Artificial Analysis)Code59.657.1−2.5
Cua Driver Bench v2 (Windows, background)Agents25.027.5+2.5
FrontierMath Tiers 1-3 (v2) - PrivateMath91.288.8−2.5
APEX-Agents 1.1Professional Work81.383.7+2.4
MMLU ProGeneral QA93.491.0−2.4
Artificial Analysis Coding Agent Index v1.5Indexes66.068.4+2.4
CursorBench 4.0Code57.855.5−2.3
Visa Agentic Commerce BenchmarkAgents62.264.4+2.2
LMArena Agent Arena Steering BurdenAgents18.614.5−2.0
SAFE-Teen 106: Urgent Safety (no prompt)Safety12.714.7−2.0
ARC-AGI-1 Semi-PrivateGeneral QA98.596.5−2.0
DAYJOB: Data ScienceProfessional Work37.335.3−2.0
DualEntry Accounting AI BenchmarkFinance & Accounting82.280.2−2.0
Terminal-Bench 2.1 Extended (Mercor held-out)Code47.545.5−2.0
BrowseComp (Mercor 130-question subset)Search88.586.6−1.9
DAYJOB: LawLegal25.127.0+1.9
Epoch Capabilities Index (ECI)Indexes169166−1.9
Harvey LAB-AA v1.0 (criterion pass rate)Legal91.293.1+1.8
SAFE-Teen 106: Health Advice (no prompt)Safety35.033.2+1.8
DeepSWE v1.1Code72.370.5−1.8
GDPval (Mercor 213-task rubric score)Professional Work88.386.5−1.8
MMLU-Bio (SecureBio)Science92.791.0−1.7
RedlineBenchLegal42.040.3−1.7
Roboflow Vision Evals (September 2026 protocol)Multimodal85.583.8−1.7
DTBenchGeneral QA98.997.3−1.6
DAYJOB: TaxFinance & Accounting9.210.8+1.6
DeepResearch Bench II (Mercor 130-task set)Search53.351.7−1.6
HealthBenchHealth & Medicine64.365.9+1.6
BioSimBench 1.0Science14.516.0+1.5
SWE-bench VerifiedCode98.296.7−1.5
SAFE-Teen 106 worst branch: Health Advice (no prompt)Safety44.342.9+1.4
SWE Atlas Codebase QnA Extended (Mercor held-out)Code43.344.7+1.4
SWE Atlas - Codebase QnACode48.950.3+1.4
Humanity's Last Exam DiamondGeneral QA54.653.3−1.3
SAFE-Teen 106 worst branch: Urgent Safety (under-18 prompt)Safety10.511.8−1.3
Political Science LLM BenchmarkProfessional Work70.268.9−1.3
GDPval-AA v2.1Professional Work18661838−1.3
HLAB-Vals-1.1Legal9.210.4+1.3
AA-AnalystAgent (pass^5)Professional Work56.357.5+1.2
Financial Coaching BenchFinance & Accounting76.274.9−1.2
SAFE-Teen 106: Relationships (under-18 prompt)Safety0.41.7−1.2
GDP.xlsxProfessional Work30.329.1−1.2
APEX-SWECode67.666.4−1.2
LMArena Agent Arena Bash ReliabilityAgents15.112.7−1.2
Access to Justice BenchmarkLegal73.874.9+1.2
VariantBench V1 Clean RerunsScience50.651.7+1.1
ForecastBench Preliminary Dataset Brier IndexForecasting62.261.1−1.1
AML BenchFinance & Accounting66.467.5+1.1
Conceptual Reasoning IndexIndexes77.977.1−1.1
FrontierSWE V2 Worst@5Code52.050.9−1.1
HardGeoBench Avg DistanceMultimodal655689−1.0
DistressBenchSafety75.674.6−1.0
τ³-Banking (Mercor run, bm25_grep retrieval)Agents55.054.0−1.0
APEX-Accounting 1.1Finance & Accounting65.564.5−1.0
Veracode GenAI Code Security (Fall 2026)Cybersecurity73.072.0−1.0
RCM-Bench-1Health & Medicine90.191.1+0.9
FrontierCode 1.1 Extended - ScoreCode63.564.4+0.9
AA-Briefcase v1.1 EloAgents18011816+0.8
DAYJOB: AccountingFinance & Accounting34.435.2+0.8
VIALSMultimodal40.241.0+0.8
MMMU ProMultimodal86.685.8−0.8
FrontierCode 1.1 Extended - Pass RateCode69.269.9+0.7
HealthBench ProfessionalHealth & Medicine71.470.8−0.6
AdvancedIFChat & Writing82.883.4+0.6
micro1 PharmaBenchHealth & Medicine10.811.4+0.6
LMArena Agent Arena Confirmed SuccessAgents20.719.5−0.6
CharXiv Extended (Mercor held-out)Multimodal84.784.2−0.5
Prophet Arena Arena Score (Agentic, OpenCode)Forecasting86.285.7−0.5
AA-LCR Extended (Mercor held-out)Search74.273.8−0.4
SAFE-Teen 106: Health Advice (under-18 prompt)Safety17.316.9+0.4
MedScribeHealth & Medicine91.491.1−0.3
CharXiv (descriptive and reasoning combined)Multimodal96.496.1−0.3
MissionReportProfessional Work63.563.8+0.3
CritPt (avg@5)Science31.731.4−0.3
InferenceBenchCode12.112.1−0.2
Excel Modeling BenchmarkFinance & Accounting75.975.7−0.2
SAFE-Teen 106: Urgent Safety (under-18 prompt)Safety3.12.9+0.2
DeepFinanceFinance & Accounting54.854.6−0.2
Prophet Arena Arena Score (Fixed Context)Forecasting85.585.3−0.2
AA-3D v1.0 EloCode12021204+0.2
FrontierSWE V2 Best@5Code71.571.7+0.2
DAYJOB: FinanceFinance & Accounting23.924.0+0.1
LMArena Agent Arena Tool HallucinationAgents0.70.9+0.1
Excel Modeling Benchmark v1.1Finance & Accounting75.875.9+0.1
FrontierMath ErdősMath2.92.90.0
LMCAGeneral QA68.268.30.0
LMArena Agent Arena Bash RecoveryAgents15.015.00.0
ActiveVisionMultimodal41.241.20.0
AIME 2025Math100.0100.00.0
ClaimBenchFinance & Accounting100.0100.00.0
Cua Driver Bench v2 (Linux, background)Agents12.512.50.0
Cua Driver Bench v2 (macOS, background)Agents57.557.50.0
Dual-Arm Sim (task completion)Agents0.00.00.0
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 - Corporate M&ALegal0.00.00.0
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 - 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
ObviousBenchGeneral QA100.0100.00.0
OTIS Mock AIME 2024-2025Math100.0100.00.0
ProofBench v1.1Math100.0100.00.0
SpatialBench V1 Clean RerunsScience72.872.80.0
sudo L7Code41.741.70.0

Claude Opus 5.5vsClaude Sonnet 5.5normalised score marginsData Nov 15, 2026Data: Vals AI, Mercor, Artificial Analysis and 86 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…