Atlas

Compare Models

GPT-5.5 vs GPT-6 Sol

GPT-6 Sol leads by 3.8 capability points (≈95% interval +1.6 to +5.9): a clear lead over GPT-5.5 · P(higher) ≈ 100%

Wins 127 of 178 shared benchmarks (46 lost, 5 tied)

Capability vs. relative cost
GPT-5.5GPT-6 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 Solcapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins46 – 127 · 178 shared, 5 tied

Comparing matching setups only (18 benchmarks excluded) · suites counted once
← GPT-5.5GPT-6 Sol →

GPT-5.5vsGPT-6 Solwins by categoryData Nov 15, 2026Data: Mercor, Vals AI, Artificial Analysis and 47 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is GPT-6 Sol minus GPT-5.5.

BrowseComp Extended (Mercor held-out)Search19.063.2+44.2
Vending-Bench 2Games752414428+41.8
BrokenArXiv 05/2026Math50.087.0+37.0
VoxelBench Text RatingGames19252394+33.4
SRE Bench — Fully SolvedCode3.830.5+26.7
BrokenArXiv 06/2026Math69.495.4+25.9
CadQueryEval (Corrected Geometry)Code76.0100.0+24.0
Olam Social Poker EloGames15211542+22.8
ChartographyMultimodal31.053.6+22.6
LMArena Code Arena WebDevCode15121673+21.1
DeepsecBench — Revalidated V1Code21.140.9+19.8
DeepsecBench — Revalidated V1 (233 issues)Code21.040.7+19.7
GameBench 2Games59.979.4+19.5
ALE-BenchCode21082637+18.7
GeneBench-Pro Public Release SubsetScience8.026.0+18.0
Vibe Code Bench v1.1Code69.887.8+18.0
FrontierMath Tier 4 (v2) - PrivateMath72.590.0+17.5
Humanity's Last Exam DiamondGeneral QA27.945.2+17.3
AutomationBench 1.0.6 (Zapier)Agents16.633.2+16.6
Terminal-Bench 4.0 (Artificial Analysis)Code14.630.3+15.7
BuseyBench SVGMultimodal6.07.5+15.3
LMArena Agent Arena Net ImprovementAgents4.28.8+15.0
AutomationBench-AA 1.0.6Agents47.361.7+14.4
Creative Writing v3 EloChat & Writing18452126+14.2
Furniture Assembly BenchmarkMultimodal44.258.3+14.2
BullshitBench v2 LeaderboardSafety45.059.0+14.0
SWE Atlas Codebase QnA Extended (Mercor held-out)Code17.331.3+14.0
LiveBench 2.0Agents68.781.9+13.2
ArXivMath 05/2026Math77.590.0+12.5
AA-Briefcase v1.1 EloAgents11351359+12.3
Code MigrationCode45.257.2+12.0
Finance Agent v2 - General Qualitative AnalysisFinance & Accounting65.653.6−11.9
ChessBench CoherenceGames100.088.2−11.8
RuneBenchGames5.76.4+11.8
BrowseComp (Mercor 130-question subset)Search77.789.4+11.7
HealthBench Hard (updated implementation)Health & Medicine33.822.1−11.7
Vals RSI Index v1.1 (four tasks)Agents16.528.1+11.6
ZeroBench Main Mean pass@1 (five samples)Multimodal17.228.8+11.6
Legal Research BenchLegal40.428.8−11.5
HealthBench (updated implementation)Health & Medicine58.447.1−11.3
BioSecBench-Function V1 Clean RerunsScience28.339.3+11.0
Roboflow Vision Evals (October 2026, five tasks)Multimodal66.076.9+10.9
RiemannBenchMath55.264.8+9.6
KernelBench Mega Normalized (RTX PRO 6000)Code12.22.7−9.5
Artificial Analysis Intelligence Index v4.3.2Indexes38.447.6+9.3
TxBench Oligo Discovery V1 Clean RerunsScience31.940.6+8.6
APEX-SWECode36.645.0+8.4
YC-BenchGames12060451441793+8.2
EnterpriseBench: CoreCraftProfessional Work51.343.1−8.2
VariantBench V1 Clean RerunsScience30.538.7+8.2
MMMU-Pro Extended (Mercor held-out)Multimodal43.851.9+8.1
NoCoT-Bench NCRI 15.2General QA100115+7.9
GBENCH One-shot Coding GScoreGames62.270.1+7.9
TxBench Preclinical Pharmacology V1 Clean RerunsScience55.363.0+7.7
Roboflow Vision Evals (September 2026 protocol)Multimodal74.882.3+7.5
VulcanBench Frontier v4Code78.585.9+7.5
Tax Agent BenchFinance & Accounting60.553.0−7.4
ArXivMath 06/2026Math83.691.0+7.3
FrontierCode 2.0 - Pass RateCode26.433.6+7.3
MedXpertQA MM Extended (Mercor held-out)Health & Medicine60.868.0+7.2
Agents' Last Exam (ALE)Agents24.531.6+7.1
TxBench Antibody Discovery V1 Clean RerunsScience29.937.0+7.0
Chess Puzzles (task version 2.0)Games63.056.0−7.0
Excel Modeling BenchmarkFinance & Accounting64.571.5+7.0
FrontierCode 2.0 - ScoreCode23.630.4+6.8
ScBench-Long V1 Clean RerunsScience21.628.4+6.8
SAGEMath51.544.8−6.7
Finance Agent v2 - PrecedentsFinance & Accounting21.928.5+6.6
EpiBench V1 Clean RerunsScience28.322.0−6.3
Extended NYT Connections 940 (Quadratic v1)Games96.290.1−6.1
SUPERChemScience66.872.8+6.0
ZeroBench Main Questions pass^5Multimodal11.017.0+6.0
Design Arena Overall Code CategoriesCode12571293+6.0
NoCoT-Bench NCKI (kspine v3)General QA111119+5.5
CritPt (avg@5)Science25.430.9+5.4
OncologyBench Strict (25 tasks, October 9 rubric)Health & Medicine14.720.0+5.3
Vals Index v2Indexes57.462.6+5.2
Conceptual Reasoning IndexIndexes67.371.0+5.1
LisanBench Path LengthGames98458222−5.0
Artificial Analysis Progress Index v1Indexes320334+4.9
MedScribeHealth & Medicine86.982.0−4.8
LMCAGeneral QA54.359.1+4.8
APEX-1 (Mercor, mean score)Professional Work68.463.6−4.8
GDPval-AA v2.1Professional Work13571466+4.8
Longform Creative WritingChat & Writing78.282.8+4.6
MedXpertQA MM (200-question set)Health & Medicine76.781.3+4.6
ARC-AGI-2General QA85.089.6+4.6
FrontierMath Tiers 1-3 (v2) - PrivateMath85.389.8+4.6
ScBench V1 Clean RerunsScience57.161.5+4.4
Terminal-Bench 2.1 Extended (Mercor held-out)Code38.042.4+4.4
LMArena Text Arena EloChat & Writing14811457−4.3
Public Benefits Bench v1.1Professional Work60.956.6−4.3
GDPval Extended (Mercor held-out rubric score)Professional Work54.558.7+4.2
SimpleBenchGeneral QA69.073.1+4.1
TaxCalcBench TY25Finance & Accounting24.028.0+4.0
SWE-bench MultilingualCode77.781.7+4.0
VCT v2 (SecureBio)Science45.349.3+4.0
MBCT (SecureBio)Science57.161.1+4.0
LMArena Agent Arena SteerabilityAgents6.914.5+3.8
Finance Agent v2 - General Quantitative AnalysisFinance & Accounting73.770.0−3.7
Finance Agent v2 - Financial ModelingFinance & Accounting22.919.3−3.6
ITBench-AAAgents45.849.4+3.6
Finance Agent v2 - All PassFinance & Accounting39.636.0−3.6
SpatialBench V1 Clean RerunsScience64.167.5+3.5
APEX-Accounting 1.1Finance & Accounting56.359.7+3.4
AA-Omniscience IndexIndexes20.527.1+3.3
HealthBench (length-adjusted)Health & Medicine56.553.2−3.3
Finance Agent v2 - Disclosure AnalysisFinance & Accounting58.955.6−3.3
HieroglyphBenchMultimodal22.826.0+3.2
LAB-Bench ProtocolQA (SecureBio)Science77.981.1+3.2
Epoch Capabilities Index (ECI)Indexes160164+3.2
Finance Agent v2 - ComparablesFinance & Accounting43.740.8−2.8
Vectara Hallucination LeaderboardSafety9.36.5+2.8
ChessBench EloGames13491303−2.7
Bio Capabilities Index (SecureBio)Indexes158161+2.7
GDP.pdf (Artificial Analysis, All-pass)Professional Work22.625.2+2.6
DeepResearch Bench II (Mercor 130-task set)Search45.442.9−2.5
Finance Agent v2 - AdjustmentsFinance & Accounting45.743.2−2.5
FrontierCode 1.1 Extended - ScoreCode56.759.1+2.5
Finance Agent v2 - Market AnalysisFinance & Accounting60.358.0−2.3
SimpleQA Verified (Epoch anti-abstention)General QA63.060.7−2.3
HealthBench ProfessionalHealth & Medicine57.259.5+2.3
DeepSWE v1.1Code64.466.6+2.2
LisanBench Difficulty-WeightedGames33163118−2.1
Harvey's Legal Agent BenchmarkLegal3.81.7−2.1
MedCodeHealth & Medicine49.147.1−2.0
FrontierCode 1.1 Extended - Pass RateCode62.864.8+2.0
LAB-Bench LitQA2 (SecureBio)Science49.851.8+2.0
ACCoRDGeneral QA80.582.4+1.9
VCT (SecureBio)Science52.354.0+1.7
Medical Long Context Reasoning (MLCR-AA)Health & Medicine17.816.1−1.7
APEX-Agents 1.1Professional Work70.368.7−1.6
SciCode 1.0.1 (Artificial Analysis)Code56.157.6+1.5
ChessBench AccuracyGames78.877.4−1.5
GPQA-Bio (SecureBio)Science87.488.9+1.5
ProgramBenchCode0.52.0+1.5
SWE-bench VerifiedCode78.880.3+1.5
FrontierMath ErdősMath0.01.5+1.5
Ruby LLM Benchmarks — Program FixerCode77.679.0+1.5
HealthBench Hard (length-adjusted)Health & Medicine31.530.1−1.4
MMMU ProMultimodal81.282.5+1.4
DTBenchGeneral QA96.097.3+1.3
ComplexConstraintsChat & Writing49.550.8+1.3
LMArena Agent Arena Bash RecoveryAgents9.57.0−1.3
Humanity's Last Exam (Text-Only)General QA45.046.3+1.3
LAB-Bench CloningScenarios (SecureBio)Science97.398.5+1.2
Terminal-Bench 2.1Code84.385.4+1.1
SWE Atlas - Codebase QnACode45.244.1−1.1
CharXiv Extended (Mercor held-out)Multimodal82.783.8+1.1
GDPval (Mercor 213-task rubric score)Professional Work84.683.5−1.1
HPCT (SecureBio)Science66.867.8+1.0
WCB (SecureBio)Science59.760.5+0.8
Creative Writing v3 Rubric ScoreChat & Writing15.916.1+0.8
Finance Agent v2 - Earnings AnalysisFinance & Accounting73.272.4−0.8
Blueprint-Bench 2Multimodal0.40.4+0.7
ForecastBench Preliminary Dataset Brier IndexForecasting63.362.6−0.7
ObviousBenchGeneral QA100.099.3−0.7
AA-LCR v1.1Search84.383.7−0.7
BioSecBench-Surveillance V1 Clean RerunsScience47.147.7+0.7
SpatialBench-Long V1 Clean RerunsScience32.131.6−0.6
SM BenchSafety65.466.0+0.5
LLM Chess EloGames15321545+0.5
ARC-AGI-1 Semi-PrivateGeneral QA95.095.5+0.5
ARC-AGI-3 (Semi-Private)General QA0.40.9+0.5
CharXiv (descriptive and reasoning combined)Multimodal94.695.0+0.4
PubMedQA (SecureBio)Health & Medicine80.580.9+0.4
SciCode (main problems)Science12.312.7+0.4
AdvancedIFChat & Writing81.380.9−0.4
AA-LCR Extended (Mercor held-out)Search70.570.2−0.3
SpeciEval (Refusal-Aware Retries, Historical)Safety98.298.5+0.3
SpeciEval (Pooled English Runs)Safety98.298.5+0.3
GlobeBenchGeneral QA25.225.0−0.2
HealthBench Consensus (updated implementation)Health & Medicine95.795.8+0.1
AIME 2025Math100.0100.00.0
MMLU ProGeneral QA88.988.90.0
Mystery Game PuzzlesGames56.056.00.0
OTIS Mock AIME 2024-2025Math100.0100.00.0
SWE-bench Verified Extended (Mercor held-out)Code39.339.30.0

GPT-5.5vsGPT-6 Solnormalised score marginsData Nov 15, 2026Data: Mercor, Vals AI, Artificial Analysis and 47 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…