Atlas

Compare Models

Claude Haiku 5.5 vs GPT-5.5

Claude Haiku 5.5 leads by 1.3 capability points (≈95% interval −0.7 to +3.3): statistically tied · P(higher) ≈ 89%

GPT-5.5 wins 66 of 122 shared benchmarks (55 lost, 1 tied)

Capability vs. relative cost
Claude Haiku 5.5GPT-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 5.5vsGPT-5.5capability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins55 – 66 · 122 shared, 1 tied

Comparing matching setups only (28 benchmarks excluded) · suites counted once
← Claude Haiku 5.5GPT-5.5 →

Claude Haiku 5.5vsGPT-5.5wins by categoryData Nov 15, 2026Data: Mercor, Vals AI, Artificial Analysis and 32 othersVillage LLM Atlasatlas.village.dev

Margins

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

Medical Long Context Reasoning (MLCR-AA)Health & Medicine63.317.8−45.6
SimpleQA Verified (Epoch anti-abstention)General QA23.863.0+39.2
A-Fantasia (Valid Attempts with Retries)Games72.334.7+37.7
Extended NYT Connections 940 (Quadratic v1)Games65.796.2+30.5
Roboflow Vision Evals — Detection mAP@75 (September)Multimodal54.324.0−30.3
Roboflow Vision Evals — Detection mAP@50:95 (September)Multimodal53.024.0−29.0
ChessBench EloGames8751349+28.2
FrontierMath Tier 4 (v2) - PrivateMath46.372.5+26.2
Mystery Game PuzzlesGames30.056.0+26.0
Roboflow Vision Evals — Detection mAP@50 (September)Multimodal68.244.2−24.0
ARC-AGI-2General QA61.385.0+23.7
Vending-Bench 2Games36027524+23.7
TaxCalcBench TY25Finance & Accounting2.024.0+22.0
BrowseComp (Mercor 130-question subset)Search56.377.7+21.4
AA-Briefcase v1.1 EloAgents15231135−21.3
Vibe Code Bench v1.1Code90.469.8−20.6
ChessBench CoherenceGames80.7100.0+19.3
RuneBenchGames4.55.7+18.8
APEX-SWECode55.136.6−18.5
NoCoT-Bench NCRI 15.2General QA68.3100+17.5
NoCoT-Bench NCKI (kspine v3)General QA85.9111+17.2
VoxelBench Text RatingGames16961925+16.3
MineBench — Bradley-Terry RatingGames15821846+16.0
LLM Chess EloGames11431532+16.0
Code2Video-AA EloCode11921000−15.6
GameBench 2Games75.159.9−15.1
Terminal-Bench 4.0 (Artificial Analysis)Code29.314.6−14.6
BuseyBench SVGMultimodal4.66.0+14.4
SWE-bench Verified Extended (Mercor held-out)Code53.339.3−14.0
ChessBench AccuracyGames66.978.8+12.8
MedXpertQA MM Extended (Mercor held-out)Health & Medicine48.460.8+12.4
KernelBench Mega Normalized (RTX PRO 6000)Code0.012.2+12.2
MMMU-Pro Extended (Mercor held-out)Multimodal31.943.8+11.9
GDPval Extended (Mercor held-out rubric score)Professional Work66.054.5−11.5
AutomationBench-AA 1.0.6Agents36.047.3+11.3
SpeciEval (Pooled English Runs)Safety86.998.2+11.3
DeepResearch Bench II (Mercor 130-task set)Search34.345.4+11.1
LMArena Text Arena EloChat & Writing14221481+10.3
FrontierMath Tiers 1-3 (v2) - PrivateMath75.185.3+10.2
SWE-bench VerifiedCode88.978.8−10.1
BullshitBench v2 LeaderboardSafety57.047.0−10.0
LisanBench Path LengthGames128979845−9.5
MedXpertQA MM (200-question set)Health & Medicine67.876.7+8.9
SM BenchSafety74.365.4−8.9
BrowseComp Extended (Mercor held-out)Search27.519.0−8.5
Chess Puzzles (task version 2.0)Games55.063.0+8.0
SWE-bench MultilingualCode85.177.7−7.4
GDPval-AA v2.1Professional Work15231357−7.3
LiveBench 2.0Agents75.668.7−6.9
AA-LCR Extended (Mercor held-out)Search63.870.5+6.7
GBENCH One-shot Coding GScoreGames55.662.2+6.6
LisanBench Difficulty-WeightedGames27113316+6.5
DAXBench v2Code81.988.3+6.4
Roboflow Vision Evals — Identification, Strict (September)Multimodal80.286.5+6.3
AA-LCR v1.1Search78.384.3+6.0
ARC-AGI-1 Semi-PrivateGeneral QA89.095.0+6.0
APEX-Accounting 1.1Finance & Accounting62.256.3−5.9
SAGEMath45.851.5+5.7
ExploitGymCybersecurity9.415.1+5.7
Excel Modeling BenchmarkFinance & Accounting59.064.5+5.5
Finance Agent v2 - Financial ModelingFinance & Accounting28.322.9−5.4
Finance Agent v2 - Disclosure AnalysisFinance & Accounting64.358.9−5.3
Finance Agent v2 - PrecedentsFinance & Accounting27.021.9−5.1
Artificial Analysis Intelligence Index v4.3.2Indexes43.438.4−5.0
Roboflow Vision Evals — Counting, Judged (September)Multimodal73.068.0−5.0
Roboflow Vision Evals — Counting, Strict (September)Multimodal73.068.0−5.0
AA-Omniscience IndexIndexes10.720.5+4.9
SUPERChemScience62.066.8+4.8
AutomationBench 1.0.6 (Zapier)Agents12.016.6+4.6
DeepSWE v1.1Code59.964.4+4.5
ALE-BenchCode19962108+4.0
Finance Agent v2 - AdjustmentsFinance & Accounting49.645.7−3.8
Design Arena Overall Code CategoriesCode12801257−3.8
AA-AnalystAgent (pass^5)Professional Work53.850.0−3.7
APEX-1 (Mercor, mean score)Professional Work64.768.4+3.7
Finance Agent v2 - General Qualitative AnalysisFinance & Accounting61.965.6+3.7
Roboflow Vision Evals (October 2026, five tasks)Multimodal69.666.0−3.6
RedlineBenchLegal40.644.1+3.5
SWE Atlas - Codebase QnACode48.745.2−3.5
Legal Research BenchLegal43.840.4−3.4
Furniture Assembly BenchmarkMultimodal47.544.2−3.3
Finance Agent v2 - All PassFinance & Accounting42.939.6−3.3
LMCAGeneral QA51.154.3+3.2
Blueprint-Bench 2Multimodal0.30.4+3.2
Roboflow Vision Evals — Identification, Judged (September)Multimodal86.589.6+3.1
CritPt (avg@5)Science22.625.4+2.9
Conceptual Reasoning IndexIndexes65.367.3+2.8
Finance Agent v2 - Market AnalysisFinance & Accounting63.160.3−2.8
MedScribeHealth & Medicine89.586.9−2.7
Roboflow Vision Evals — Reasoning, Judged (September)Multimodal74.872.2−2.6
ForecastBench Preliminary Dataset Brier IndexForecasting60.863.3+2.5
Harvey's Legal Agent BenchmarkLegal1.33.8+2.5
MedCodeHealth & Medicine51.549.1−2.4
ARC-AGI-3 (Semi-Private)General QA2.80.4−2.4
Humanity's Last Exam (Text-Only)General QA42.745.0+2.4
Public Benefits Bench v1.1Professional Work63.260.9−2.3
Artificial Analysis Progress Index v1Indexes326320−2.2
MMMU ProMultimodal79.181.2+2.1
APEX-Agents 1.1Professional Work68.370.3+2.0
ProgramBenchCode2.50.5−2.0
SRE Bench — Fully SolvedCode5.73.8−1.9
GDP.pdf (Artificial Analysis, All-pass)Professional Work20.822.6+1.8
Tax Agent BenchFinance & Accounting62.260.5−1.7
Humanity's Last Exam DiamondGeneral QA29.627.9−1.7
Finance Agent v2 - ComparablesFinance & Accounting45.343.7−1.7
MMLU ProGeneral QA87.488.9+1.5
SciCode 1.0.1 (Artificial Analysis)Code55.056.1+1.2
OTIS Mock AIME 2024-2025Math98.9100.0+1.1
ACCoRDGeneral QA81.580.5−1.0
DTBenchGeneral QA95.096.0+1.0
GDPval (Mercor 213-task rubric score)Professional Work85.584.6−0.9
Terminal-Bench 2.1Code85.084.3−0.7
Finance Agent v2 - General Quantitative AnalysisFinance & Accounting74.473.7−0.6
Roboflow Vision Evals — Reasoning, Strict (September)Multimodal72.872.2−0.6
Code MigrationCode44.645.2+0.6
AdvancedIFChat & Writing80.881.3+0.5
CharXiv (descriptive and reasoning combined)Multimodal94.194.6+0.5
Finance Agent v2 - Earnings AnalysisFinance & Accounting73.573.2−0.3
Terminal-Bench 2.1 Extended (Mercor held-out)Code37.738.0+0.3
CharXiv Extended (Mercor held-out)Multimodal82.982.7−0.2
Epoch Capabilities Index (ECI)Indexes160160−0.2
AIME 2025Math100.0100.00.0

Claude Haiku 5.5vsGPT-5.5normalised score marginsData Nov 15, 2026Data: Mercor, Vals AI, Artificial Analysis and 32 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…