Atlas

Compare Models

GPT-5.5 vs GPT-6.1 Sol

GPT-6.1 Sol leads by 6.3 capability points (≈95% interval +4.0 to +8.6): a clear lead over GPT-5.5 · P(higher) ≈ 100%

Wins 135 of 165 shared benchmarks (25 lost, 5 tied)

Capability vs. relative cost
GPT-5.5GPT-6.1 SolClosed 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-5.5vsGPT-6.1 Solcapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins25 – 135 · 165 shared, 5 tied

Comparing matching setups only (16 benchmarks excluded) · suites counted once
← GPT-5.5GPT-6.1 Sol →

GPT-5.5vsGPT-6.1 Solwins by categoryData Nov 15, 2026Data: Vals AI, Mercor, Artificial Analysis and 44 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is GPT-6.1 Sol minus GPT-5.5.

ActiveVisionMultimodal10.671.8+61.2
BrowseComp Extended (Mercor held-out)Search19.067.7+48.7
SRE Bench — Fully SolvedCode3.850.8+46.9
VoxelBench Text RatingGames19252550+44.5
BrokenArXiv 05/2026Math50.090.0+40.0
Clocktower BenchGames14861764+39.9
Terminal-Bench 4.0 (Artificial Analysis)Code14.654.0+39.4
Furniture Assembly BenchmarkMultimodal44.280.0+35.8
GameBench 2Games59.992.0+32.1
ALE-BenchCode21082990+31.2
Humanity's Last Exam DiamondGeneral QA27.958.8+30.9
BrokenArXiv 06/2026Math69.4100.0+30.6
LMArena Code Arena WebDevCode15121743+30.2
Code2Video-AA EloCode10001369+30.0
ClockBench AccuracyMultimodal46.174.4+28.3
FrontierMath Tier 4 (v2) - PrivateMath72.5100.0+27.5
TextQuests Average Cumulative HarmSafety21.429.3−26.9
NoCoT-Bench NCRI 15.2General QA100149+26.8
ARC-AGI-3 (Semi-Private)General QA0.426.7+26.3
GeneBench-Pro Public Release SubsetScience8.033.0+25.0
Mystery Game PuzzlesGames56.080.0+24.0
Olam Social Poker EloGames15211543+23.9
Agent Red Teaming k=100 (CAIS)Safety41.518.8+22.8
HiL-Bench (Human-in-Loop Benchmark)Agents39.762.3+22.7
TextQuests Game ProgressGames48.971.4+22.6
EnigmaEvalGeneral QA37.259.5+22.3
AA-Briefcase v1.1 EloAgents11351504+20.2
Code MigrationCode45.265.1+20.0
BuseyBench SVGMultimodal6.08.0+19.5
AutomationBench-AA 1.0.6Agents47.366.6+19.3
Vibe Code Bench v1.1Code69.888.9+19.1
TxBench Oligo Discovery V1 (113 evaluations)Science33.952.5+18.6
DeepsecBench — Revalidated V1 (233 issues)Code21.039.5+18.5
RuneBenchGames5.76.9+18.4
BullshitBench v2 LeaderboardSafety47.065.0+18.0
BioSecBench-Function V1 Clean RerunsScience28.346.0+17.7
BrowseComp (Mercor 130-question subset)Search77.795.4+17.7
ArXivMath 05/2026Math77.595.0+17.5
MMMU-Pro Extended (Mercor held-out)Multimodal43.861.3+17.5
GBENCH One-shot Coding GScoreGames62.279.0+16.8
LiveBench 2.0Agents68.785.4+16.7
TxBench Antibody Discovery V1 (100 evaluations)Science33.049.7+16.7
Design Arena Overall Code CategoriesCode12571356+16.4
Medical Long Context Reasoning (MLCR-AA)Health & Medicine17.833.9+16.1
SpatialViz-BenchMultimodal74.290.3+16.1
Roboflow Vision Evals (October 2026, five tasks)Multimodal66.081.7+15.7
VariantBench V1 Clean RerunsScience30.546.0+15.5
PostTrainBench v1.2Indexes23.938.8+14.9
SWE Atlas Codebase QnA Extended (Mercor held-out)Code17.331.3+14.0
TaxCalcBench TY25Finance & Accounting24.038.0+14.0
SimpleBenchGeneral QA69.082.9+13.9
Artificial Analysis Intelligence Index v4.3.2Indexes38.451.8+13.5
NoCoT-Bench NCKI (kspine v3)General QA111130+13.0
HPCT Refusal (CAIS)Safety29.042.0+13.0
ChessBench EloGames13491562+12.7
ChessBench CoherenceGames100.087.5−12.5
IntPhys 2Multimodal59.972.3+12.4
Agents' Last Exam Overall ScoreAgents46.958.6+11.7
SUPERChemScience66.877.8+11.0
SimpleQA Verified (Epoch anti-abstention)General QA63.073.9+10.9
DeepSWE v1.1Code64.475.2+10.8
ArXivMath 06/2026Math83.694.4+10.8
Roboflow Vision Evals (September 2026 protocol)Multimodal74.885.5+10.7
AA-Omniscience IndexIndexes20.541.5+10.5
VulcanBench Frontier v4Code78.588.8+10.3
APEX-SWECode36.646.9+10.3
Vals RSI Index v1.1 (four tasks)Agents16.526.6+10.1
Artificial Analysis Progress Index v1Indexes320348+10.0
HealthBench ProfessionalHealth & Medicine57.267.2+10.0
FrontierCode 2.0 - Pass RateCode26.436.1+9.7
YC-BenchGames1206045936267−9.4
FrontierCode 2.0 - ScoreCode23.633.0+9.4
OncologyBench Strict (25 tasks, October 9 rubric)Health & Medicine14.724.0+9.3
MindCube TinyMultimodal79.989.1+9.2
ARC-AGI-2General QA85.094.2+9.2
GDPval-AA v2.1Professional Work13571563+9.1
GDP.pdfProfessional Work26.035.0+9.0
MedXpertQA MM (200-question set)Health & Medicine76.785.7+9.0
Machiavelli Deception (CAIS)Safety81.778.8+8.7
HieroglyphBenchMultimodal22.831.4+8.6
LMCAGeneral QA54.362.8+8.5
SWE-bench VerifiedCode78.887.3+8.5
FrontierMath Tiers 1-3 (v2) - PrivateMath85.393.7+8.4
MASKSafety90.198.3+8.2
APEX-Accounting 1.1Finance & Accounting56.364.3+8.0
Terminal-Bench 2.1 Extended (Mercor held-out)Code38.045.8+7.8
SpatialBench-Long V1 Clean RerunsScience32.139.9+7.7
ERQAMultimodal70.578.2+7.7
Humanity's Last ExamGeneral QA43.851.3+7.5
LisanBench Difficulty-WeightedGames33164004+7.4
Conceptual Reasoning IndexIndexes67.372.7+7.4
SWE-bench MultilingualCode77.784.9+7.2
Epoch Capabilities Index (ECI)Indexes160169+7.1
ScBench V1 Clean RerunsScience57.163.6+6.5
CritPt (avg@5)Science25.431.7+6.3
Excel Modeling BenchmarkFinance & Accounting64.570.8+6.2
SWE-bench Verified Extended (Mercor held-out)Code39.345.3+6.0
ScBench-Long V1 Clean RerunsScience21.626.7+5.1
Chess Puzzles (task version 2.0)Games63.068.0+5.0
SAGEMath51.546.5−5.0
HealthBench Hard (length-adjusted)Health & Medicine31.536.2+4.7
FrontierMath ErdősMath0.04.4+4.4
TxBench Preclinical Pharmacology V1 Clean RerunsScience55.359.7+4.3
DAXBench v2Code88.392.6+4.3
Finance Agent v2 - Market AnalysisFinance & Accounting60.364.5+4.2
MMMU ProMultimodal81.285.1+3.9
MedXpertQA MM Extended (Mercor held-out)Health & Medicine60.864.7+3.9
Finance Agent v2 - PrecedentsFinance & Accounting21.925.8+3.9
FrontierCode 1.1 Extended - ScoreCode56.760.4+3.7
BioSecBench-Surveillance V1 Clean RerunsScience47.150.8+3.7
FrontierCode 1.1 Extended - Pass RateCode62.866.4+3.7
SM BenchSafety65.469.0+3.5
ARC-AGI-1 Semi-PrivateGeneral QA95.098.5+3.5
LisanBench Path LengthGames98458736−3.4
LLM Chess EloGames15321612+3.3
SWE Atlas - Codebase QnACode45.248.4+3.2
SpatialBench V1 Clean RerunsScience64.167.3+3.2
APEX-Agents 1.1Professional Work70.373.2+2.9
PCT Political Inconsistency (CAIS)Safety40.037.3+2.7
Ruby LLM Benchmarks — Program FixerCode77.675.0−2.6
LMArena Agent Arena Bash RecoveryAgents9.54.5−2.5
ProgramBenchCode0.53.0+2.5
CharXiv Extended (Mercor held-out)Multimodal82.785.1+2.4
Vision Arena OverallMultimodal12831293+2.4
Finance Agent v2 - General Qualitative AnalysisFinance & Accounting65.663.2−2.4
SciCode (main problems)Science12.314.6+2.3
MMLU ProGeneral QA88.991.0+2.1
ChessBench AccuracyGames78.880.7+2.0
HealthBench (length-adjusted)Health & Medicine56.558.5+2.0
Legal Research BenchLegal40.438.5−1.9
Tax Agent BenchFinance & Accounting60.562.3+1.9
KernelBench Mega Normalized (RTX PRO 6000)Code12.214.1+1.8
HealthBench (updated implementation)Health & Medicine58.456.7−1.7
Finance Agent v2 - Earnings AnalysisFinance & Accounting73.271.5−1.7
Harvey's Legal Agent BenchmarkLegal3.85.4+1.7
ForecastBench Preliminary Dataset Brier IndexForecasting63.361.7−1.6
Public Benefits Bench v1.1Professional Work60.959.3−1.6
CharXiv (descriptive and reasoning combined)Multimodal94.696.0+1.4
Terminal-Bench 2.1Code84.385.4+1.1
AdvancedIFChat & Writing81.382.4+1.1
Finance Agent v2 - ComparablesFinance & Accounting43.742.6−1.0
Finance Agent v2 - AdjustmentsFinance & Accounting45.744.7−1.0
EnterpriseBench: CoreCraftProfessional Work51.350.3−1.0
SpeciEval (Pooled English Runs)Safety98.297.4−0.8
SpeciEval (Refusal-Aware Retries, Historical)Safety98.297.4−0.8
Finance Agent v2 - General Quantitative AnalysisFinance & Accounting73.774.5+0.8
Extended NYT Connections 940 (Quadratic v1)Games96.295.5−0.7
DTBenchGeneral QA96.096.5+0.5
Finance Agent v2 - Disclosure AnalysisFinance & Accounting58.958.4−0.5
MedScribeHealth & Medicine86.986.5−0.4
HealthBench Hard (updated implementation)Health & Medicine33.833.4−0.4
ACCoRDGeneral QA80.580.9+0.4
SciCode 1.0.1 (Artificial Analysis)Code56.155.8−0.3
Finance Agent v2 - All PassFinance & Accounting39.639.9+0.3
AA-LCR v1.1Search84.384.0−0.3
MedCodeHealth & Medicine49.148.8−0.3
HealthBench Consensus (updated implementation)Health & Medicine95.795.9+0.2
Finance Agent v2 - Financial ModelingFinance & Accounting22.923.1+0.2
LMArena Text Arena EloChat & Writing14811482+0.1
VCT-Refusal (CAIS)Safety16.016.1+0.1
AA-AnalystAgent (pass^5)Professional Work50.050.00.0
AIME 2025Math100.0100.00.0
EpiBench V1 Clean RerunsScience28.328.30.0
ObviousBenchGeneral QA100.0100.00.0
OTIS Mock AIME 2024-2025Math100.0100.00.0

GPT-5.5vsGPT-6.1 Solnormalised score marginsData Nov 15, 2026Data: Vals AI, Mercor, Artificial Analysis and 44 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…