Atlas

Compare Models

Claude Haiku 5.5 vs GPT-5.6 Terra

Claude Haiku 5.5 leads by 0.6 capability points (≈95% interval −1.4 to +2.7): statistically tied · P(higher) ≈ 73%

Wins 69 of 166 shared benchmarks (68 lost, 29 tied)

Capability vs. relative cost
Claude Haiku 5.5GPT-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.

Claude Haiku 5.5vsGPT-5.6 Terracapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins69 – 68 · 166 shared, 29 tied

Comparing matching setups only (34 benchmarks excluded) · suites counted once
← Claude Haiku 5.5GPT-5.6 Terra →

Claude Haiku 5.5vsGPT-5.6 Terrawins by categoryData Nov 15, 2026Data: Vals AI, Mercor, Artificial Analysis and 30 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 Claude Haiku 5.5.

IOI (Vals v2)Code47.387.6+40.3
Medical Long Context Reasoning (MLCR-AA)Health & Medicine63.331.7−31.7
LisanBench Path LengthGames128973303−29.8
BrowseComp (Mercor 130-question subset)Search56.385.8+29.5
A-Fantasia (Valid Attempts with Retries)Games72.343.3+29.0
SWE-bench Verified Extended (Mercor held-out)Code53.378.7+25.4
FrontierMath Tier 4 (v2) - PrivateMath46.370.7+24.4
APEX-SWECode55.131.1−24.0
AutomationBench-AA 1.0.6Agents36.059.6+23.7
Vending-Bench 2Games36027343+22.7
ARC-AGI-2General QA61.383.9+22.7
RuneBenchGames4.55.9+21.9
GameBench 2Games75.153.5−21.6
HardGeoBench Avg DistanceMultimodal13191991−20.7
Harvey's Legal Agent Benchmark - Structured Finance SecuritizationLegal0.020.0+20.0
Eval Connections One-Shot with TrapsGames67.086.0+19.0
GDPval Extended (Mercor held-out rubric score)Professional Work66.047.2−18.8
LisanBench Difficulty-WeightedGames27111197−16.2
Vibe Code Bench v1.1Code90.474.6−15.9
MMMU-Pro Extended (Mercor held-out)Multimodal31.947.4+15.5
SM BenchSafety74.359.0−15.3
BuseyBench SVGMultimodal4.66.1+15.2
ProofBench v1.1Math89.074.0−15.0
Roboflow Vision Evals — Detection mAP@75 (September)Multimodal54.340.6−13.7
Terminal-Bench 4.0Code36.422.7−13.6
Roboflow Vision Evals — Detection mAP@50:95 (September)Multimodal53.039.6−13.4
AA-Briefcase v1.1 EloAgents15731333−13.1
DeepResearch Bench II (Mercor 130-task set)Search34.347.2+12.9
Extended NYT Connections 940 (Quadratic v1)Games65.778.4+12.7
NoCoT-Bench NCKI (kspine v3)General QA85.9104+12.3
NoCoT-Bench NCRI 15.2General QA68.389.0+11.3
SpeciEval (Pooled English Runs)Safety86.998.1+11.1
SWE-bench VerifiedCode88.977.8−11.1
FrontierCode 2.0 (Artificial Analysis)Code34.923.9−11.0
ChessBench CoherenceGames80.769.7−11.0
FrontierMath Tiers 1-3 (v2) - PrivateMath75.186.0+10.9
AA-LCR Extended (Mercor held-out)Search63.874.4+10.6
Roboflow Vision Evals — Reasoning, Judged (September)Multimodal74.865.3−9.5
SWE-bench MultilingualCode85.194.4+9.3
HardGeoBench <=100miMultimodal45.036.3−8.7
MedCodeHealth & Medicine51.543.4−8.1
Chess Puzzles (task version 2.0)Games55.047.0−8.0
LMArena Text Arena EloChat & Writing14221468+7.9
MedXpertQA MM (200-question set)Health & Medicine67.875.7+7.9
GBENCH One-shot Coding GScoreGames55.663.3+7.7
ARC-AGI-1 Semi-PrivateGeneral QA89.096.5+7.5
Roboflow Vision Evals — Reasoning, Strict (September)Multimodal72.865.3−7.5
CritPt (avg@5)Science22.630.0+7.4
Wiz Cyber Model Arena v3.9Cybersecurity57.064.4+7.4
GDPval-AA v2.1Professional Work16291461−7.4
Roboflow Vision Evals — Counting, Judged (September)Multimodal73.065.8−7.2
Roboflow Vision Evals — Counting, Strict (September)Multimodal73.065.8−7.2
RedlineBenchLegal40.647.8+7.2
Excel Modeling BenchmarkFinance & Accounting59.066.2+7.2
CursorBench 4.0Code48.441.3−7.1
Design Arena Overall Code CategoriesCode12801238−7.0
Roboflow Vision Evals — Detection mAP@50 (September)Multimodal68.261.3−6.9
ACCoRDGeneral QA81.574.7−6.9
MedScribeHealth & Medicine89.582.9−6.7
Furniture Assembly BenchmarkMultimodal47.554.2+6.7
Harvey HLAB — Criteria Pass RateLegal73.479.8+6.3
HardGeoBench Country AccuracyMultimodal51.345.0−6.2
Terminal-Bench-Science 0.1Science20.013.8−6.2
LLM Chess EloGames11431293+6.2
CVDP (517-problem set)Code57.050.9−6.1
Finance Agent v2 - General Qualitative AnalysisFinance & Accounting61.967.8+5.9
CritPt v2 (agentic, avg@5)Science64.069.7+5.7
ALE-BenchCode19962154+5.6
ChessBench AccuracyGames66.971.9+5.4
AA-Omniscience IndexIndexes10.70.1−5.3
Finance Agent v2 - Financial ModelingFinance & Accounting28.323.0−5.3
SWE-Bench Pro V2 — FullCode87.292.4+5.1
HardGeoBench <=10miMultimodal30.025.0−5.0
Mystery Game PuzzlesGames30.035.0+5.0
APEX-1 (Mercor, mean score)Professional Work64.769.5+4.8
GDPval (Mercor 213-task rubric score)Professional Work85.580.7−4.8
DistressBenchSafety71.776.4+4.8
LMArena Agent Arena Confirmed SuccessAgents1.6-7.2−4.4
LMArena Agent Arena Net Improvement (Steering Burden)Agents0.5-1.8−4.4
τ³-Banking (Mercor run, bm25_grep retrieval)Agents43.038.8−4.2
Artificial Analysis Intelligence Index v5.0.1Indexes36.532.3−4.1
LMCAGeneral QA51.155.0+3.9
SRE Bench — Fully SolvedCode5.79.5+3.8
GDP.pdf (Artificial Analysis, All-pass)Professional Work20.824.6+3.8
Finance Agent v2 - Earnings AnalysisFinance & Accounting73.577.3+3.8
Roboflow Vision Evals (October 2026, five tasks)Multimodal69.666.0−3.6
APEX-Agents 1.1Professional Work68.371.8+3.5
SWE Atlas - Codebase QnACode48.745.2−3.5
CyberBench v1.1 OverallCybersecurity75.572.1−3.4
LMArena Agent Arena Bash ReliabilityAgents7.40.7−3.3
Code MigrationCode44.647.8+3.2
Terminal-Bench 2.1Code85.088.0+3.0
BullshitBench v2 LeaderboardSafety57.054.0−3.0
Tax Agent BenchFinance & Accounting62.265.2+3.0
ChessBench EloGames875925+3.0
SUPERChemScience62.064.9+2.9
SAGE v1.1Math89.887.0−2.8
AutomationBench 1.0.6 (Zapier)Agents12.014.8+2.7
AdvancedIFChat & Writing80.878.1−2.7
Finance Agent v2 - Disclosure AnalysisFinance & Accounting64.366.9+2.7
Political Science LLM BenchmarkProfessional Work65.768.3+2.6
MedXpertQA MM Extended (Mercor held-out)Health & Medicine48.451.0+2.6
Harvey LAB (Mercor run, undisclosed ~213-task cohort)Legal4.92.3−2.6
Terminal-Bench 4.0 (Artificial Analysis)Code32.835.4+2.5
Finance Agent v2 - Market AnalysisFinance & Accounting63.160.6−2.5
Legal Research BenchLegal43.841.3−2.4
APEX-Accounting 1.1Finance & Accounting62.259.8−2.4
BrowseComp Extended (Mercor held-out)Search27.525.3−2.2
Blueprint-Bench 2Multimodal0.30.3−2.2
Finance Agent v2 - PrecedentsFinance & Accounting27.024.8−2.1
ProgramBenchCode2.50.5−2.0
CharXiv Extended (Mercor held-out)Multimodal82.981.1−1.8
Finance Agent v2 - ComparablesFinance & Accounting45.347.0+1.7
Finance Agent v2 - General Quantitative AnalysisFinance & Accounting74.472.7−1.7
DTBenchGeneral QA95.093.3−1.7
LMArena Agent Arena Tool HallucinationAgents-2.11.0+1.5
LMArena Agent Arena Steering BurdenAgents2.4-0.5−1.4
Vals Index v2.1Indexes54.453.1−1.3
Artificial Analysis Intelligence Index v4.3.2Indexes43.442.1−1.3
Terminal-Bench 2.1 Extended (Mercor held-out)Code37.736.4−1.3
SAGEMath45.847.0+1.2
OTIS Mock AIME 2024-2025Math98.999.7+0.8
Public Benefits Bench v1.1Professional Work63.262.4−0.8
Humanity's Last Exam DiamondGeneral QA29.628.8−0.8
Artificial Analysis Progress Index v1Indexes326324−0.8
Humanity's Last Exam (Text-Only)General QA42.741.9−0.8
LMArena Agent Arena Praise vs ComplaintAgents-3.5-1.9+0.8
MMLU ProGeneral QA87.488.1+0.7
AA-LCR v1.1Search78.379.0+0.7
LiveBench 2.0Agents75.675.0−0.7
Conceptual Reasoning IndexIndexes65.365.7+0.6
Epoch Capabilities Index (ECI)Indexes160161+0.5
MMMU ProMultimodal79.179.5+0.4
ARC-AGI-3 (Semi-Private)General QA0.50.8+0.3
DeepSWE v1.1Code59.960.2+0.3
Finance Agent v2 - AdjustmentsFinance & Accounting49.649.7+0.2
Finance Agent v2 - All PassFinance & Accounting42.942.8−0.1
AIME 2025Math100.0100.00.0
CharXiv (descriptive and reasoning combined)Multimodal94.194.10.0
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 - 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 - Trusts Estates Private ClientLegal0.00.00.0
Harvey's Legal Agent Benchmark - White Collar Defense InvestigationsLegal0.00.00.0
Roboflow Vision Evals — Identification, Judged (September)Multimodal86.586.50.0
Roboflow Vision Evals — Identification, Strict (September)Multimodal80.280.20.0
SciCode 1.0.1 (Artificial Analysis)Code55.055.00.0
Terminal-Bench Science 0.1 (Vals)Science10.010.00.0

Claude Haiku 5.5vsGPT-5.6 Terranormalised score marginsData Nov 15, 2026Data: Vals AI, Mercor, Artificial Analysis and 30 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…