Atlas

Compare Models

Claude Haiku 5.5 vs Claude Opus 5.5

Claude Opus 5.5 leads by 6.4 capability points (≈95% interval +4.0 to +8.8): a clear lead over Claude Haiku 5.5 · P(higher) ≈ 100%

Wins 170 of 201 shared benchmarks (6 lost, 25 tied)

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

Benchmark wins6 – 170 · 201 shared, 25 tied

Comparing matching setups only (29 benchmarks excluded) · suites counted once
← Claude Haiku 5.5Claude Opus 5.5 →

Claude Haiku 5.5vsClaude Opus 5.5wins by categoryData Nov 15, 2026Data: Vals AI, Mercor, Artificial Analysis and 55 othersVillage LLM Atlasatlas.village.dev

Margins

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

KernelBench Mega Normalized (RTX PRO 6000)Code0.0100.0+100.0
InferenceBenchCode4.412.1+91.3
VoxelBench Text RatingGames16962613+65.3
Toloka Arena (v3)Agents21.075.0+54.0
FrontierMath Tier 4 (v2) - PrivateMath46.395.0+48.7
SimpleQA Verified (Epoch anti-abstention)General QA23.872.2+48.4
Terminal-Bench-Science 0.1Science18.161.9+43.8
Mystery Game PuzzlesGames30.071.0+41.0
TaxCalcBench TY25Finance & Accounting2.042.0+40.0
AA-PersonalityBench Organization indexChat & Writing41.079.1+38.5
LMArena Agent Arena Net Improvement (Steering Burden)Agents0.519.5+36.0
Furniture Assembly BenchmarkMultimodal47.583.3+35.8
LisanBench Difficulty-WeightedGames27115963+34.8
KernelBench CUDA Normalized (RTX PRO 6000)Code41.476.3+34.8
LisanBench Path LengthGames1289724029+34.6
Decision-Bench v1.1Professional Work25.359.1+33.9
MineBench — Bradley-Terry RatingGames15822134+33.4
BrowseComp Extended (Mercor held-out)Search27.560.5+33.0
BrowseComp (Mercor 130-question subset)Search56.388.5+32.2
ARC-AGI-2General QA61.393.3+32.1
Terminal-Bench 4.0Code33.364.8+31.6
AA-PersonalityBench Imagination indexChat & Writing63.894.3+30.8
Terminal-Bench 4.0 (Artificial Analysis)Code29.359.6+30.3
AA-3D v1.0 EloCode9981202+29.7
Artificial Analysis Coding Agent Index v1.5Indexes36.866.0+29.2
LMArena Code Arena WebDevCode15941814+28.9
ChessBench EloGames8751349+28.2
Code2Video-AA EloCode11921536+27.9
EBR-bench v4 (card ban, single agent)Games43.871.4+27.6
ALE-BenchCode19962748+26.6
SWE-bench Multimodal (Vals v1, 480 Tasks)Code44.070.2+26.3
ExploitGymCybersecurity9.434.5+25.1
Humanity's Last Exam DiamondGeneral QA29.654.6+25.0
ChessBenchGames41.365.3+24.0
FrontierML ($200 wallet)Agents34.958.8+23.9
FrontierPhysics v0.1Science4.227.4+23.2
BuseyBench SVGMultimodal4.66.8+22.9
MedXpertQA MM Extended (Mercor held-out)Health & Medicine48.471.2+22.8
Extended NYT Connections 940 (Quadratic v1)Games65.788.5+22.8
LMArena Agent Arena Praise vs ComplaintAgents-3.540.3+21.9
HardGeoBench Avg DistanceMultimodal1319669+20.1
Harvey's Legal Agent Benchmark - Energy Natural ResourcesLegal0.020.0+20.0
Harvey's Legal Agent Benchmark - Healthcare Life SciencesLegal0.020.0+20.0
FrontierCode 2.0 - Pass RateCode31.951.8+20.0
ChartographyMultimodal46.466.3+19.9
GameBench 2Games75.194.8+19.7
WeirdML v3Science11.731.2+19.5
Bug Hunt BenchCode20.539.7+19.2
MMMU-Pro Extended (Mercor held-out)Multimodal31.951.0+19.1
Eval Connections One-Shot with TrapsGames67.086.0+19.0
DeepResearch Bench II (Mercor 130-task set)Search34.353.3+19.0
FrontierSWE V2 Best@5Code52.871.5+18.8
FrontierCode 2.0 - ScoreCode28.947.3+18.4
GBENCH One-shot Coding GScoreGames55.673.9+18.3
Blueprint-Bench 2Multimodal0.30.5+18.2
SWE Atlas v1.1 - OverallCode59.477.4+18.0
Agent Initiative BenchmarkAgents6.07.8+17.9
MedXpertQA MM (200-question set)Health & Medicine67.885.7+17.9
AA-Omniscience IndexIndexes10.746.4+17.9
LLM Chess EloGames11431575+17.7
SUPERChemScience62.079.6+17.6
FrontierSWE V2 Worst@5Code34.552.0+17.5
Conceptual Reasoning IndexIndexes65.377.9+17.3
Design Arena Overall Code CategoriesCode12801384+17.2
LMCAGeneral QA51.168.2+17.2
ProgramBench (Mercor 100-task subset)Code3.020.0+17.0
Excel Modeling BenchmarkFinance & Accounting59.075.9+16.9
HardGeoBench <=100miMultimodal45.061.3+16.3
HardGeoBench Country AccuracyMultimodal51.367.5+16.3
FrontierMath Tiers 1-3 (v2) - PrivateMath75.191.2+16.1
o11y-bench 1.1 Pass^3Agents78.193.8+15.6
AA-PersonalityBench Informativeness indexChat & Writing75.390.5+15.4
AA-PersonalityBench Warmth indexChat & Writing69.584.8+15.4
CritPt v2 (agentic, avg@5)Science66.381.4+15.1
Humanity's Last Exam (Text-Only)General QA42.757.5+14.8
OSWorld 2.1 offline - reported partial reward, step limit unspecifiedAgents72.487.2+14.8
DeepFinanceFinance & Accounting40.054.8+14.8
LMArena Text Arena EloChat & Writing14221506+14.6
DAXBench v2Code81.995.7+13.8
AA-PersonalityBench Expressiveness indexChat & Writing77.290.5+13.5
AA-PersonalityBench Transparency indexChat & Writing77.290.5+13.5
ChessBench AccuracyGames66.979.3+13.3
Roboflow Vision Evals — Reasoning, Strict (September)Multimodal72.885.9+13.1
APEX-Agents 1.1Professional Work68.381.3+13.0
SWE-bench Verified Extended (Mercor held-out)Code53.366.0+12.7
AA-Briefcase v1.1 EloAgents15731801+12.5
HardGeoBench <=10miMultimodal30.042.5+12.5
APEX-SWECode55.167.6+12.5
Cua Driver Bench v2 (macOS, background)Agents45.057.5+12.5
Roboflow Vision Evals (October 2026, five tasks)Multimodal69.682.0+12.4
DeepSWE v1.1Code59.972.3+12.4
τ³-Banking (Mercor run, bm25_grep retrieval)Agents43.055.0+12.0
Cua Driver Bench v2 (macOS, foreground)Agents76.088.0+12.0
Chess Puzzles (task version 2.0)Games55.067.0+12.0
SciCode 1.0.1 (Artificial Analysis)Code55.066.9+11.9
Roboflow Vision Evals — Reasoning, Judged (September)Multimodal74.885.9+11.1
ProofBench v1.1Math89.0100.0+11.0
LiveBench 2.0Agents75.686.5+10.9
GDPval-AA v2.1Professional Work16291866+10.5
SWE-bench MultilingualCode85.195.5+10.4
AA-LCR Extended (Mercor held-out)Search63.874.2+10.4
Excel Modeling Benchmark v1.1Finance & Accounting65.575.8+10.4
Immigration BenchLegal61.872.1+10.3
GDPval Extended (Mercor held-out rubric score)Professional Work66.075.8+9.8
Terminal-Bench 2.1 Extended (Mercor held-out)Code37.747.5+9.8
AA-PersonalityBench Slop indexChat & Writing10.520.0+9.6
LMArena Agent Arena Confirmed SuccessAgents1.620.7+9.5
ARC-AGI-1 Semi-PrivateGeneral QA89.098.5+9.5
CursorBench 4.0Code48.457.8+9.4
Roboflow Vision Evals — Identification, Judged (September)Multimodal86.595.8+9.3
SWE-bench VerifiedCode88.998.2+9.3
CritPt (avg@5)Science22.631.7+9.1
Roboflow Vision Evals — Counting, Judged (September)Multimodal73.082.0+9.0
Ramp Accounting BenchFinance & Accounting42.350.5+8.2
LMArena Agent Arena Steering BurdenAgents2.418.6+8.1
Roboflow Vision Evals — Counting, Strict (September)Multimodal73.081.1+8.1
GDP.pdf (Artificial Analysis, All-pass)Professional Work20.828.8+8.0
Cua Driver Bench v2 (Linux, foreground)Agents68.076.0+8.0
Roboflow Vision Evals — Detection mAP@50 (September)Multimodal68.276.0+7.8
MMMU ProMultimodal79.186.6+7.5
SM BenchSafety74.381.8+7.4
Roboflow Vision Evals — Identification, Strict (September)Multimodal80.287.5+7.3
CVDP (517-problem set)Code57.064.1+7.1
GDPxProfessional Work4.912.0+7.1
ACCoRDGeneral QA81.588.4+6.9
Epoch Capabilities Index (ECI)Indexes160169+6.8
HealthBench ProfessionalHealth & Medicine64.871.4+6.7
Roboflow Vision Evals — Detection mAP@75 (September)Multimodal54.360.9+6.6
AA-LCR v1.1Search78.384.7+6.3
BullshitBench v2 LeaderboardSafety57.063.0+6.0
DualEntry Accounting AI BenchmarkFinance & Accounting76.282.2+6.0
MMLU ProGeneral QA87.493.4+6.0
Harvey LAB (Mercor run, undisclosed ~213-task cohort)Legal4.910.8+5.9
Hermes BenchAgents71.076.7+5.8
Roboflow Vision Evals — Detection mAP@50:95 (September)Multimodal53.058.3+5.3
FrontierCode 1.1 Extended - ScoreCode58.463.6+5.1
Cua Driver Bench v2 (Windows, background)Agents20.025.0+5.0
HLAB-Vals-1.1Legal4.29.2+5.0
FrontierCode 1.1 Extended - Pass RateCode64.769.5+4.8
Finance Agent v2Finance & Accounting54.158.6+4.5
Cua Driver Bench v2 (Windows, foreground)Agents70.074.0+4.0
DTBenchGeneral QA95.098.9+3.9
DistressBenchSafety71.775.6+3.9
LMArena Agent Arena Bash ReliabilityAgents7.415.1+3.9
SWE-Marathon v1.1Code50.046.3−3.7
FORTRESS Average Risk Score (ARS)Safety9.513.2−3.7
Medical Long Context Reasoning (MLCR-AA)Health & Medicine63.366.7+3.3
APEX-Accounting 1.1Finance & Accounting62.265.5+3.3
HealthBenchHealth & Medicine61.264.3+3.1
GDPval (Mercor 213-task rubric score)Professional Work85.588.3+2.8
AA-AnalystAgent (pass^5)Professional Work53.856.3+2.5
Harvey HLAB — Criteria Pass RateLegal73.475.8+2.3
CharXiv (descriptive and reasoning combined)Multimodal94.196.4+2.3
AdvancedIFChat & Writing80.882.8+2.0
VulcanBench Frontier v4Code84.486.4+1.9
AA-PersonalityBench Conviction indexChat & Writing67.669.5+1.9
MedScribeHealth & Medicine89.591.4+1.9
CharXiv Extended (Mercor held-out)Multimodal82.984.7+1.8
RCM-Bench-1Health & Medicine88.490.1+1.8
MedCodeHealth & Medicine51.549.8−1.7
Terminal-Bench 2.1Code85.083.5−1.5
BioSimBench 1.0Science13.014.5+1.5
HealthBench (length-adjusted, Artificial Analysis)Health & Medicine58.557.0−1.5
SAGE v1.1Math89.891.3+1.5
ForecastBench Preliminary Dataset Brier IndexForecasting60.862.2+1.4
RedlineBenchLegal40.642.0+1.4
LMArena Agent Arena Tool HallucinationAgents-2.10.7+1.4
Harvey LAB-AA v1.0 (criterion pass rate)Legal89.991.2+1.3
OTIS Mock AIME 2024-2025Math98.9100.0+1.1
APEX-1 (Mercor, mean score)Professional Work64.765.8+1.1
ChessBench CoherenceGames80.780.1−0.6
Prophet Arena Arena Score (Agentic, OpenCode)Forecasting85.986.2+0.3
SpeciEval (Pooled English Runs)Safety86.987.2+0.3
SWE Atlas - Codebase QnACode48.748.9+0.2
Prophet Arena Arena Score (Fixed Context)Forecasting85.485.5+0.1
SAGEMath45.845.80.0
AIME 2025Math100.0100.00.0
ClaimBenchFinance & Accounting100.0100.00.0
Cua Driver Bench v2 (Linux, background)Agents12.512.50.0
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 - 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 - Structured Finance SecuritizationLegal0.00.00.0
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

Claude Haiku 5.5vsClaude Opus 5.5normalised score marginsData Nov 15, 2026Data: Vals AI, Mercor, Artificial Analysis and 55 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…