Atlas

Compare Models

GPT-5.5 vs Grok 4.6

GPT-5.5 leads by 0.0 capability points (≈95% interval −1.9 to +2.0): statistically tied · P(higher) ≈ 51%

Wins 109 of 218 shared benchmarks (90 lost, 19 tied)

Capability vs. relative cost
GPT-5.5Grok 4.6Closed 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.

GPT-5.5vsGrok 4.6capability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins109 – 90 · 218 shared, 19 tied

Comparing matching setups only (71 benchmarks excluded) · suites counted once
← GPT-5.5Grok 4.6 →

GPT-5.5vsGrok 4.6wins by categoryData Nov 15, 2026Data: Vals AI, Mercor, benchmarks.bio and 50 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Grok 4.6 minus GPT-5.5.

KernelBench Hard Normalized (RTX PRO 6000)Code0.060.5+60.5
LLM Creative Story WritingChat & Writing2.4-2.9−60.1
LLM Chess EloGames1532443−44.7
Document ArenaProfessional Work14841438−41.3
FrontierMath Tier 4 (v2) - PrivateMath72.531.7−40.8
SWE-bench Verified Extended (Mercor held-out)Code39.378.0+38.7
MRCR v2 (8-needle, Context Arena) - 128K pointwiseGeneral QA85.852.6−33.2
MirrorCode (ML, +Private, 2L)Code10.041.1+31.1
ChessBench CoherenceGames100.069.2−30.8
Chess Puzzles (task version 2.0)Games63.033.0−30.0
ChessBench EloGames1349891−27.2
BioSecBench-RefusalSafety37.164.2+27.1
Olam Social Poker EloGames15211499−23.9
AA-Briefcase EloAgents11361545+23.8
SM BenchSafety65.489.0+23.6
SAGEMath51.528.9−22.6
Mystery Game PuzzlesGames56.034.0−22.0
APEX-SWECode36.657.7+21.1
LAB-Bench LitQA2 (SecureBio)Science49.828.9−20.9
Harvey's Legal Agent Benchmark - Antitrust CompetitionLegal0.020.0+20.0
Harvey's Legal Agent Benchmark - Banking FinanceLegal0.020.0+20.0
Harvey's Legal Agent Benchmark - Corporate GovernanceLegal0.020.0+20.0
Harvey's Legal Agent Benchmark - Corporate M&ALegal0.020.0+20.0
Harvey's Legal Agent Benchmark - Energy Natural ResourcesLegal20.040.0+20.0
Harvey's Legal Agent Benchmark - Funds Asset ManagementLegal0.020.0+20.0
Harvey's Legal Agent Benchmark - Healthcare Life SciencesLegal0.020.0+20.0
Harvey's Legal Agent Benchmark - Litigation Dispute ResolutionLegal0.020.0+20.0
Harvey's Legal Agent Benchmark - Structured Finance SecuritizationLegal0.020.0+20.0
AutomationBench-AA 1.0.6Agents47.367.0+19.7
FrontierMath Tiers 1-3 (v2) - PrivateMath85.366.0−19.3
MRCR v2 (8-needle, Context Arena) - 128K AUCGeneral QA91.773.4−18.3
TxBench Antibody Discovery V0Science33.651.9+18.3
ARC-AGI-2General QA85.067.1−17.9
MMMU-Pro Extended (Mercor held-out)Multimodal43.826.0−17.8
WeirdML (v2)Science84.967.3−17.6
Antidote: Everyday EditionSafety1013966−16.8
Extended NYT Connections 940 (Quadratic v1)Games96.280.0−16.2
BioSecBench-Function V1 Clean RerunsScience28.344.0+15.7
Bio Capabilities Index (SecureBio)Indexes158137−15.2
SWE-bench MultilingualCode77.792.7+15.0
TxBench Antibody Discovery V1 Clean RerunsScience29.944.5+14.6
Hemingway-benchChat & Writing1035983−14.4
EnterpriseBench: CoreCraftProfessional Work51.365.6+14.3
BuseyBench SVGMultimodal6.04.6−13.9
SimpleQA Verified (Epoch anti-abstention)General QA63.049.3−13.7
SpatialBench-Long V0 VerifiedScience33.319.7−13.6
DeepSWE v1.1Code54.067.5+13.5
GDPval-AA v2Professional Work13721643+13.3
GDPval-AA v2.1Professional Work13571658+13.3
ComplexConstraintsChat & Writing49.536.5−13.0
SWE-bench VerifiedCode82.695.6+13.0
YC-BenchGames12060451568241+12.6
Harvey HLAB — Criteria Pass RateLegal80.092.5+12.5
MedXpertQA MM Extended (Mercor held-out)Health & Medicine60.848.4−12.4
KernelBench Mega Normalized (RTX PRO 6000)Code12.20.0−12.2
Finance Agent v2 - General Qualitative AnalysisFinance & Accounting65.677.8+12.2
ChessBench AccuracyGames78.867.8−11.8
𝜏³-BankingAgents39.050.7+11.8
SkillsBench without SkillsAgents49.938.2−11.7
CursorBench 3.2Code58.469.9+11.5
TxBench Oligo Discovery V1 Clean RerunsScience31.943.3+11.4
Tax Agent BenchFinance & Accounting60.571.4+10.9
LLM Debate BenchmarkChat & Writing15281604+10.9
EpiBenchScience45.034.3−10.7
ChartographyMultimodal31.020.5−10.5
TxBench Oligo Discovery V0Science33.644.0+10.3
LMArena Agent Arena Net ImprovementAgents4.21.1−10.3
ALE-BenchCode21081822−10.1
VariantBench V1 Clean RerunsScience30.540.4+9.9
LAB-Bench ProtocolQA (SecureBio)Science77.968.1−9.8
WCB (SecureBio)Science59.750.3−9.4
Vending-Bench 2Games75249047+9.2
VoxelBench Text RatingGames19252050+8.9
ACCoRDGeneral QA80.571.6−8.9
GDP.pdfProfessional Work26.017.2−8.8
GameBench 2Games59.968.7+8.8
AA-AnalystAgent (pass^5)Professional Work50.041.3−8.7
VariantBenchScience29.738.1+8.4
WordleBenchGames97.089.0−8.0
Legal Research BenchLegal40.448.1+7.7
ARC-AGI-1 Semi-PrivateGeneral QA95.087.5−7.5
Terminal-Bench 2.1 Extended (Mercor held-out)Code38.045.5+7.5
NoCoT-Bench NCRI 15.2General QA10086.8−7.4
SpeciEval (Pooled English Runs)Safety98.290.8−7.4
SpeciEval (Refusal-Aware Retries, Historical)Safety98.290.8−7.4
GDPval Extended (Mercor held-out rubric score)Professional Work54.561.8+7.3
APEX-Agents 1.1Professional Work70.377.3+7.0
GeneBench-Pro Public Release SubsetScience8.015.0+7.0
Roboflow Vision Evals (October 2026, five tasks)Multimodal66.059.1−6.9
SimpleBenchGeneral QA69.075.9+6.9
GBENCH One-shot Coding GScoreGames62.255.4−6.9
MBCT (SecureBio)Science57.150.4−6.7
Conceptual Reasoning IndexIndexes67.362.5−6.6
SkillsBench with SkillsAgents62.255.8−6.4
Humanity's Last Exam DiamondGeneral QA27.921.5−6.4
Vibe Code Bench v1.1Code69.876.2+6.4
TxBench Preclinical Pharmacology V1 Clean RerunsScience55.361.7+6.3
BrowseComp (Mercor 130-question subset)Search77.784.0+6.3
LiveBench 2.0Agents68.774.9+6.2
ZeroBench Main Mean pass@1 (five samples)Multimodal17.211.0−6.2
Roboflow Vision Evals (September 2026 protocol)Multimodal74.868.7−6.1
Artificial Analysis Intelligence Index v4.3.2Indexes38.444.3+6.0
Public Benefits Bench v1.1Professional Work60.966.8+6.0
SpatialBench-Long V1 Clean RerunsScience32.126.2−5.9
LMCAGeneral QA54.348.5−5.8
HANDBOOK.mdProfessional Work21.527.3+5.8
Artificial Analysis Intelligence Index v4.3Indexes38.644.4+5.8
CritPt (avg@5)Science25.419.7−5.7
Medical Long Context Reasoning (MLCR-AA)Health & Medicine17.812.2−5.6
BioSecBench-Function V0Science36.542.0+5.5
Finance Agent v2 - Disclosure AnalysisFinance & Accounting58.964.4+5.4
CharXiv Extended (Mercor held-out)Multimodal82.777.3−5.4
AA-LCR Extended (Mercor held-out)Search70.565.2−5.3
HPCT (SecureBio)Science66.861.6−5.2
Vision Arena OverallMultimodal12831263−5.2
Design Arena Overall Code CategoriesCode12571288+5.1
ScBench-Long V1 Clean RerunsScience21.616.5−5.1
SWE Atlas - Codebase QnACode45.250.3+5.1
AA-Omniscience IndexIndexes20.530.5+5.0
DeepsecBench — Revalidated V1Code21.116.2−5.0
DeepsecBench — Revalidated V1 (233 issues)Code21.016.1−4.9
EpiBench V1 Clean RerunsScience28.323.6−4.7
SpatialBench V0 VerifiedScience65.269.9+4.6
MortgageTax (Vals all 1,258 documents)Finance & Accounting68.864.2−4.6
RedlineBenchLegal44.148.5+4.4
MedCodeHealth & Medicine49.144.7−4.4
LMArena Text Arena EloChat & Writing14811456−4.3
Furniture Assembly BenchmarkMultimodal44.240.0−4.2
Finance Agent v2 - Financial ModelingFinance & Accounting22.918.9−4.0
VCT (SecureBio)Science52.348.3−4.0
OncologyBench Strict (25 tasks, October 9 rubric)Health & Medicine14.718.7+4.0
APEX-Accounting 1.1Finance & Accounting56.360.3+4.0
BabaIsBench Stage 0: The IntroGames100.096.0−4.0
CadQueryEval (Corrected Geometry)Code76.072.0−4.0
ZeroBench Main Questions pass^5Multimodal11.07.0−4.0
TaxEval v2Finance & Accounting75.071.1−3.9
Finance Agent v2 - AdjustmentsFinance & Accounting45.749.6+3.9
CorpFin v2 - Max Fitting ContextFinance & Accounting70.066.2−3.8
EBR-bench v3Games34.330.5−3.8
Terminal-Bench 2.1Code84.388.0+3.7
Finance Agent v2 - All PassFinance & Accounting39.642.9+3.3
AA-LCR v1.1Search84.381.0−3.3
Senior SWE-BenchCode29.526.3−3.2
SpatialBench V1 Clean RerunsScience64.167.3+3.2
LAB-Bench CloningScenarios (SecureBio)Science97.394.2−3.1
Blueprint-Bench 2Multimodal0.40.3−3.0
GlobeBenchGeneral QA25.228.1+2.9
LiveCodeBench (Vals AI public dataset)Code85.388.2+2.9
Toloka Arena (v1)Agents52.349.7−2.6
Terminal-Bench 4.0 (Artificial Analysis)Code14.617.2+2.5
BrowseComp Extended (Mercor held-out)Search19.021.5+2.5
HieroglyphBenchMultimodal22.820.3−2.5
Time Horizon Index: KSPAgents8.35.8−2.5
Finance Agent v2 - ComparablesFinance & Accounting43.741.2−2.4
FoodTruck BenchGames6140859537−2.4
SciCode (main problems)Science12.314.6+2.3
ForecastBench Preliminary Dataset Brier IndexForecasting63.361.0−2.3
DAXBench v2Code88.386.2−2.1
MineBench — Bradley-Terry RatingGames18461811−2.1
CyberGymCybersecurity81.879.7−2.1
ScBench V1 Clean RerunsScience57.159.1+2.1
SWE Atlas Codebase QnA Extended (Mercor held-out)Code17.319.3+2.0
Finance Agent v2 - Market AnalysisFinance & Accounting60.362.2+2.0
LMArena Agent Arena Bash RecoveryAgents9.55.7−1.9
Excel Modeling BenchmarkFinance & Accounting64.562.7−1.8
LMArena Agent Arena SteerabilityAgents6.93.3−1.8
Vals Index v2Indexes57.459.2+1.8
CorpFin v2 - Shared Max ContextFinance & Accounting69.667.8−1.7
Epoch Capabilities Index (ECI)Indexes160158−1.7
EnterpriseOps-Gym-AAAgents46.648.3+1.7
CharXiv (descriptive and reasoning combined)Multimodal94.692.9−1.7
ARC-AGI-3 (Semi-Private)General QA0.42.1+1.7
Humanity's Last ExamGeneral QA41.439.7−1.7
ScBench-Long V0 VerifiedScience22.724.2+1.5
BioSecBench-SurveillanceScience50.251.7+1.5
MedXpertQA MM (200-question set)Health & Medicine76.775.2−1.5
GDPval (Mercor 213-task rubric score)Professional Work84.686.0+1.4
DTBenchGeneral QA96.097.3+1.3
ForecastBench Tournament Overall Brier IndexForecasting67.065.8−1.2
CorpFin v2 - Exact PagesFinance & Accounting65.664.5−1.2
SUPERChemScience66.865.7−1.1
OTIS Mock AIME 2024-2025Math100.099.2−0.8
Finance Agent v2 - Earnings AnalysisFinance & Accounting73.273.9+0.7
APEX-1 (Mercor, mean score)Professional Work68.467.7−0.7
NoCoT-Bench NCKI (kspine v3)General QA111112+0.7
MMMU ProMultimodal81.280.5−0.7
Code MigrationCode45.244.6−0.6
PG-LLM-50Science0.30.3−0.5
Artificial Analysis Progress Index v1Indexes320322+0.5
MMLU ProGeneral QA88.989.4+0.5
ProgramBenchCode0.51.0+0.5
Finance Agent v2 - General Quantitative AnalysisFinance & Accounting73.773.2−0.5
BioSecBench-Surveillance V1 Clean RerunsScience47.147.5+0.4
AdvancedIFChat & Writing81.380.9−0.4
SciCode 1.0.1 (Artificial Analysis)Code56.156.5+0.3
MedScribeHealth & Medicine86.986.5−0.3
LegalBenchLegal86.586.3−0.2
GPQA DiamondGeneral QA95.595.50.0
Finance Agent v2 - PrecedentsFinance & Accounting21.921.90.0
AHK-EvalCode97.297.20.0
AIME 2025Math100.0100.00.0
GPQA-Bio (SecureBio)Science87.487.40.0
Harvey's Legal Agent Benchmark - Arbitration International Dispute ResolutionLegal0.00.00.0
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 - Data Privacy CybersecurityLegal20.020.00.0
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 - 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 - 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
ObviousBenchGeneral QA100.0100.00.0

GPT-5.5vsGrok 4.6normalised score marginsData Nov 15, 2026Data: Vals AI, Mercor, benchmarks.bio and 50 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…