Atlas

Compare Models

GPT-5.5 vs GPT-5.6 Sol

GPT-5.6 Sol leads by 2.8 capability points (≈95% interval +0.8 to +4.9): a clear lead over GPT-5.5 · P(higher) ≈ 100%

Wins 246 of 337 shared benchmarks (82 lost, 9 tied)

Capability vs. relative cost
GPT-5.5GPT-5.6 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-5.6 Solcapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins82 – 246 · 337 shared, 9 tied

Comparing matching setups only (41 benchmarks excluded) · suites counted once
← GPT-5.5GPT-5.6 Sol →

GPT-5.5vsGPT-5.6 Solwins by categoryData Nov 15, 2026Data: Vals AI, Artificial Analysis, benchmarks.bio and 98 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is GPT-5.6 Sol minus GPT-5.5.

KernelBench Hard Normalized (H100)Code0.048.6+48.7
NanoGPT Speedrun Frontier (human record gap closed)Agents8.135.9+37.4
Voratiq Coding Agent LeaderboardCode18892299+37.0
GSO Opt@1Code40.276.5+36.3
SnakeBench RatingGames37.424.6−34.6
TextQuests Average Cumulative HarmSafety21.431.4−33.9
GraphWalks BFS - 1MGeneral QA45.477.1+31.7
ExploitBench - release-reported exploit progressCybersecurity47.978.5+30.6
ProofBenchMath50.077.0+27.0
SRE Bench — Fully SolvedCode3.830.5+26.7
VoxelBench Text RatingGames19252297+26.5
SnakeBench Best ApplesGames26.018.0−23.5
ScBench-LongScience15.938.1+22.2
ClockBench AccuracyMultimodal46.168.3+22.2
ScBench-Long V0 VerifiedScience22.743.9+21.2
LisanBench Difficulty-WeightedGames33161403−20.5
GameBench 2Games59.980.1+20.1
RiemannBenchMath55.274.4+19.2
EBR-bench v2Games21.038.6+17.6
AA-Briefcase EloAgents11361435+17.4
GraphWalks BFS - 256KGeneral QA73.790.7+17.0
Opus Magnum CampaignGames53.970.9+17.0
LLM Debate BenchmarkChat & Writing15281645+16.7
YC-BenchGames1206045728001−16.7
LisanBench Path LengthGames98454699−16.0
EQ-Bench 4Chat & Writing12951236−15.5
Time Horizon Index: KSPAgents8.323.8+15.5
Aikido Known-CVE Detection (Pass@3 Union)Cybersecurity73.188.5+15.4
OSWorld 2.0 500-step Partial ScoreAgents47.562.6+15.1
GeneBench-ProScience12.026.8+14.8
DeepsecBenchCode21.235.6+14.4
DeepsecBench — Revalidated V1Code21.135.4+14.3
DeepsecBench — Revalidated V1 (233 issues)Code21.035.3+14.3
Olam Social Poker EloGames15211534+14.1
ChartographyMultimodal31.045.0+14.0
BankerToolBench v1.1Finance & Accounting15.029.0+14.0
Clocktower BenchGames14861583+13.9
SWE-bench VerifiedCode82.696.2+13.6
CyberBench PoCCybersecurity79.793.2+13.6
LMArena Code Arena WebDevCode15121615+13.5
ChessBench — Historical Output RepairGames29.342.7+13.3
BrowseComp Extended (Mercor held-out)Search19.032.3+13.3
ActiveVisionMultimodal10.623.5+12.9
Vending-Bench 2Games75249619+12.7
Furniture Assembly BenchmarkMultimodal44.256.7+12.5
VariantBenchScience29.742.1+12.4
Drone-BenchAgents63.976.1+12.2
AutomationBench 1.0.6 (Zapier)Agents16.628.8+12.2
CadQueryEvalCode64.076.0+12.0
CadQueryEval (Corrected Geometry)Code76.088.0+12.0
SnakeBench Total ApplesGames18136.0−11.0
ErdosBenchMath2.73.1+11.0
ChessBench CoherenceGames100.089.1−10.9
MLS Bench LiteScience35.546.2+10.7
ALE-BenchCode21082410+10.7
Vibe Code Bench v1.1Code69.880.5+10.6
FoodTruck BenchGames6140853229−10.5
FrontierMath Tier 4 (v2) - PrivateMath72.582.9+10.4
ITBench-AAAgents45.856.2+10.4
RSI-Exam 0.1 (Full 88)Agents0.30.4+10.2
Terminal-Bench 4.0 (Artificial Analysis)Code14.624.7+10.1
MirrorCode (ML, +Private, 2L)Code10.020.0+10.0
OpenAI MRCR v2 8-needle - 256K-512KGeneral QA81.591.5+10.0
SnakeBench Win RateGames90.0100.0+10.0
GDPval-AA v2 (normalized)Professional Work49.559.5+9.9
GDPval-AA v2.1Professional Work13571577+9.7
KernelBench Mega — RTX PRO 6000 Best Decode SpeedupCode4.32.6−9.5
HPCT (SecureBio)Science66.876.3+9.5
SpatialBench V0 VerifiedScience65.274.5+9.3
APEX-SWECode36.645.8+9.2
Artificial Analysis Agentic IndexIndexes44.954.0+9.1
τ²-bench (Telecom)Agents93.984.8−9.1
PostTrainBench Average (v1.1)Indexes27.236.2+9.0
CursorBench 3.2Code58.467.2+8.8
Artificial Analysis Intelligence Index v4.3.2Indexes38.447.0+8.6
OpenAI Capture-the-Flag challenges - internal release scoreCybersecurity88.196.7+8.6
Artificial Analysis Intelligence Index v4.3Indexes38.647.1+8.4
Medical Long Context Reasoning (MLCR-AA)Health & Medicine17.826.1+8.3
Bio Capabilities Index (SecureBio)Indexes158169+8.1
SkillsBench with SkillsAgents62.254.1−8.1
VulcanBench Frontier v4Code78.586.5+8.0
ScBench-Long V1 Clean RerunsScience21.629.6+8.0
SnakeBench Average ApplesGames16.518.0+7.8
Excel Modeling BenchmarkFinance & Accounting64.572.3+7.8
ChessBench EloGames13491218−7.8
Code MigrationCode45.252.9+7.8
Legal Research BenchLegal40.448.1+7.7
CyberBench OverallCybersecurity80.588.1+7.6
LiveBench 2.0Agents68.776.3+7.6
ARC-AGI-2General QA85.092.5+7.5
Tax Agent BenchFinance & Accounting60.568.0+7.5
GDPval-AA v2Professional Work13721524+7.5
Vals RSI Index v1.1 (four tasks)Agents16.523.9+7.4
HiL-Bench (Human-in-Loop Benchmark)Agents39.732.3−7.3
PMPP-HardCode52.259.4+7.2
BioSecBench-Function V1 Clean RerunsScience28.335.5+7.2
OEIS Open Lite ($200)Math36.043.0+7.0
HealthBench ProfessionalHealth & Medicine57.264.1+6.9
NoCoT-Bench NCKI (kspine v3)General QA111121+6.8
BabaIsBench Stage 1: The LakeGames86.793.3+6.7
LMArena Agent Arena Net ImprovementAgents4.26.3+6.7
MRCR v2 (8-needle, Context Arena) - 128K pointwiseGeneral QA85.892.4+6.6
SimpleQA VerifiedGeneral QA63.169.7+6.6
SkillsBench without SkillsAgents49.943.3−6.6
SpatialViz-BenchMultimodal74.280.7+6.5
Creative Writing v3 EloChat & Writing18451974+6.5
DECK-BenchProfessional Work68.274.7+6.5
UnityBench-Mini3DCode66.272.5+6.3
Diligence Stack Agent BenchFinance & Accounting82.088.3+6.3
Vals Index v2Indexes57.463.7+6.3
RedlineBenchLegal44.150.3+6.2
Agents' Last Exam (ALE)Agents24.530.6+6.1
DeepResearch Bench II (Mercor 130-task set)Search45.451.5+6.1
Hemingway-benchChat & Writing10351057+6.1
BrowseCompSearch84.490.4+6.0
SWE-Bench Pro (Public Dataset)Code58.664.6+6.0
TaxCalcBench TY25Finance & Accounting24.030.0+6.0
TxBench Preclinical Pharmacology V1 Clean RerunsScience55.361.3+6.0
HardGeoBench Avg DistanceMultimodal11961381−5.7
Public Benefits Bench v1.1Professional Work60.966.5+5.6
OSWorld-VerifiedAgents77.583.0+5.5
GBENCH One-shot Coding GScoreGames62.267.7+5.4
MRCR v2 (8-needle, Context Arena) - 128K AUCGeneral QA91.797.1+5.4
PointerbenchMultimodal59.965.3+5.4
𝜏³-BankingAgents39.044.3+5.4
SpatialBench-Long V1 Clean RerunsScience32.137.5+5.4
BabyVision with PythonMultimodal83.688.9+5.3
BabyVision with toolsMultimodal83.688.9+5.3
MASKSafety90.195.4+5.3
DeepSearchQA (F1 score)Search87.893.1+5.3
TasteVal (compute multiplier)Science0.50.6+5.3
Terminal-Bench 2.1Code84.389.5+5.2
Vals RSI Index v1.1 (historical five tasks)Agents17.923.1+5.2
Senior SWE-BenchCode29.534.7+5.2
AutomationBench (Overall)Agents12.918.1+5.2
Vals IndexIndexes68.073.1+5.2
PolicyBench US Exact MatchFinance & Accounting83.588.7+5.1
MedCodeHealth & Medicine49.144.0−5.1
EnterpriseBench: CoreCraftProfessional Work51.346.2−5.1
NoCoT-Bench NCRI 15.2General QA100109+5.1
DeepSWE v1.1Code64.469.4+5.0
Antidote: Everyday EditionSafety10131027+5.0
Surface Evolver BenchAgents88.193.1+5.0
HPCT Refusal (CAIS)Safety29.024.0−5.0
ArXivMath 06/2026Math83.688.5+4.9
DrugDiscoveryBenchScience51.656.5+4.9
CorpFin v2 - Max Fitting ContextFinance & Accounting70.065.2−4.9
HieroglyphBenchMultimodal22.827.6+4.8
ZeroBench Main Mean pass@1 (five samples)Multimodal17.222.0+4.8
VariantBench V1 Clean RerunsScience30.535.3+4.8
VCT-Refusal (CAIS)Safety16.011.3−4.7
WCB (SecureBio)Science59.764.4+4.7
Toloka Arena (v1)Agents52.347.6−4.7
GDP.pdfProfessional Work26.030.7+4.7
ACCoRDGeneral QA80.575.9−4.6
Vals Multimodal IndexIndexes68.172.6+4.6
KernelBench Mega Normalized (H100)Code23.118.6−4.5
Finance Agent v2 - Disclosure AnalysisFinance & Accounting58.963.3+4.4
LMArena Agent Arena Bash RecoveryAgents9.50.7−4.4
SM BenchSafety65.461.1−4.3
Finance Agent v2 - PrecedentsFinance & Accounting21.926.2+4.3
LMCAGeneral QA54.358.5+4.2
Kimi Code Bench v2Code69.064.8−4.2
SUPERChemScience66.871.0+4.2
VCT v2 (SecureBio)Science45.349.4+4.1
CorpFin v2 - Exact PagesFinance & Accounting65.661.5−4.1
SpatialBench V1 Clean RerunsScience64.168.1+4.1
IntPhys 2Multimodal59.963.9+4.0
Artificial Analysis Progress Index v1Indexes320331+4.0
OncologyBench Strict (25 tasks, October 9 rubric)Health & Medicine14.718.7+4.0
Chess Puzzles (task version 2.0)Games63.059.0−4.0
LAB-Bench ProtocolQA (SecureBio)Science77.981.9+4.0
MBCT (SecureBio)Science57.161.1+4.0
PerceptionBenchMultimodal55.859.7+3.9
SciCode (main problems)Science12.316.2+3.9
Finance Agent v2 - Market AnalysisFinance & Accounting60.364.2+3.9
FrontierMath Tiers 1-3 (v2) - PrivateMath85.389.1+3.9
WeirdML (v2)Science84.988.8+3.8
Document ArenaProfessional Work14841480−3.8
EnterpriseOps-Gym-AAAgents46.642.9−3.7
OmniDocBench v1.5Multimodal89.485.8−3.6
MATH-VisionMultimodal92.295.8+3.6
Longform Creative WritingChat & Writing78.281.7+3.5
Finance Agent v2 - AdjustmentsFinance & Accounting45.749.2+3.5
ObviousBenchGeneral QA100.096.5−3.5
Artificial Analysis Coding IndexIndexes74.978.3+3.5
FrontierFinance v3.1Finance & Accounting43.546.8+3.4
SpreadsheetBench 2Professional Work29.132.4+3.4
RuneBenchGames5.75.9+3.3
FrontierCode 1.1 Extended - ScoreCode56.760.0+3.3
SWE Atlas - Test WritingCode42.645.9+3.3
APEX-AccountingFinance & Accounting48.251.5+3.3
WorldVQA ForceAnswerMultimodal38.541.8+3.3
PG-LLM-50Science0.30.4+3.3
FrontierCode 1.1 Extended - Pass RateCode62.866.0+3.3
TextQuests Game ProgressGames48.952.1+3.2
IFBench (Artificial Analysis avg@5)Chat & Writing75.972.7−3.2
CorpFin v2 - Shared Max ContextFinance & Accounting69.666.4−3.1
CorpFin v2 (Vals Index component)Finance & Accounting69.666.4−3.1
CorpFin v2 (Vals Multimodal Index component)Finance & Accounting69.666.4−3.1
CritPt (avg@5)Science25.428.6+3.1
Finance Agent v2 (Vals Index subset)Finance & Accounting51.855.0+3.1
Finance Agent v2 (Vals Multimodal Index component)Finance & Accounting51.855.0+3.1
Vectara Hallucination LeaderboardSafety9.312.4−3.1
MindCube TinyMultimodal79.976.8−3.1
TxBench Oligo Discovery V1 Clean RerunsScience31.935.0+3.1
SpatialBench-Long V0 VerifiedScience33.336.4+3.0
Machiavelli Deception (CAIS)Safety81.782.7−3.0
LAB-Bench LitQA2 (SecureBio)Science49.852.8+3.0
MMLU Pro - EngineeringGeneral QA84.887.8+3.0
CliniCARE-BenchHealth & Medicine76.173.2−2.9
EpiBenchScience45.042.1−2.8
HealthBench (updated implementation)Health & Medicine58.455.6−2.8
AHK-EvalCode97.294.4−2.8
ScBench V1 Clean RerunsScience57.159.8+2.7
CyberGymCybersecurity81.884.5+2.7
HealthBench Hard (updated implementation)Health & Medicine33.831.1−2.7
LiveCodeBench (Vals AI public dataset)Code85.382.6−2.7
Blueprint-Bench 2Multimodal0.40.3−2.6
ChessBench AccuracyGames78.876.4−2.6
EpiBench V1 Clean RerunsScience28.325.8−2.5
AA-AnalystAgent (pass^5)Professional Work50.047.5−2.5
ARC-AGI-1 Semi-PrivateGeneral QA95.097.5+2.5
CRED v0.92Science96.599.0+2.5
DiligenceBenchFinance & Accounting47.550.0+2.5
HardGeoBench Country AccuracyMultimodal45.047.5+2.5
Humanity's Last Exam DiamondGeneral QA27.930.4+2.5
HardGeoBench <=10miMultimodal28.826.3−2.5
Ruby LLM Benchmarks — Program FixerCode77.675.2−2.4
BioSecBench-RefusalSafety37.139.5+2.4
ToolathlonAgents55.658.0+2.4
Epoch Capabilities Index (ECI)Indexes160163+2.4
AA-LCRSearch73.371.0−2.3
OfficeQA ProProfessional Work60.963.2+2.3
BioSecBench-SurveillanceScience50.252.5+2.3
Terminal-Bench HardCode59.862.1+2.3
Finance Agent v2 - Financial ModelingFinance & Accounting22.920.7−2.2
MMLU Pro - LawGeneral QA76.979.1+2.2
BrokenArXiv 06/2026Math69.467.3−2.2
GPQA-Bio (SecureBio)Science87.485.3−2.1
HANDBOOK.mdProfessional Work21.523.5+2.0
ZeroBench Main Questions pass^5Multimodal11.013.0+2.0
Mystery Game PuzzlesGames56.058.0+2.0
VCT (SecureBio)Science52.354.3+2.0
Conceptual Reasoning IndexIndexes67.368.7+1.9
ForecastBench Preliminary Dataset Brier IndexForecasting63.361.4−1.9
Finance Agent v2 - General Quantitative AnalysisFinance & Accounting73.775.6+1.9
Roboflow Vision EvalsMultimodal76.478.3+1.8
Finance Agent v2 - All PassFinance & Accounting39.641.4+1.8
BioSecBench-Function V0Science36.538.3+1.8
Humanity's Last ExamGeneral QA43.845.5+1.7
ARC-AGI-3 (Semi-Private)General QA0.42.1+1.7
PCT Political Inconsistency (CAIS)Safety40.041.7−1.7
CyberBench PatchCybersecurity81.483.1+1.7
MedScribeHealth & Medicine86.985.2−1.6
Finance Agent v2 - ComparablesFinance & Accounting43.745.3+1.6
HealthBench Hard (length-adjusted)Health & Medicine31.533.1+1.6
GlobeBenchGeneral QA25.223.7−1.5
FrontierCode 2.0 - Pass RateCode26.427.9+1.5
BioTIER RefusalSafety73.471.9−1.5
LAB-Bench CloningScenarios (SecureBio)Science97.398.8+1.5
MortgageTax (Vals all 1,258 documents)Finance & Accounting68.867.3−1.5
TxBench Antibody Discovery V1 Clean RerunsScience29.931.4+1.5
Creative Writing v3 Rubric ScoreChat & Writing15.916.2+1.4
Extended NYT Connections 940-Puzzle SetGames97.596.1−1.4
FrontierCode 2.0 - ScoreCode23.625.0+1.4
MMMU ProMultimodal83.284.6+1.4
A-Fantasia (Valid Attempts with Retries)Games34.736.0−1.3
A-FantasiaGeneral QA34.736.0−1.3
MMLU Pro - BusinessGeneral QA92.093.3+1.3
Harvey's Legal Agent BenchmarkLegal3.82.5−1.2
MMLU Pro - PhysicsGeneral QA91.592.8+1.2
Agent Red Teaming k=100 (CAIS)Safety41.540.3+1.2
Vision Arena OverallMultimodal12831288+1.2
Frontier Financial JudgementFinance & Accounting52.451.2−1.2
Finance Agent v2 - Earnings AnalysisFinance & Accounting73.274.4+1.2
BuseyBench SVGMultimodal6.06.1+1.1
SciCode (Artificial Analysis avg@3)Science55.956.9+1.0
SAGEMath51.552.6+1.0
BullshitBench v2 LeaderboardSafety47.048.0+1.0
ComplexConstraintsChat & Writing49.550.5+1.0
ProgramBenchCode0.51.5+1.0
MMLU Pro - Computer ScienceGeneral QA89.890.7+1.0
MMLU Pro (Vals 5-Shot CoT)General QA88.189.1+1.0
ForecastBench Tournament Overall Brier IndexForecasting67.066.1−0.9
TxBench Oligo Discovery V0Science33.634.5+0.9
MMLU Pro - OtherGeneral QA86.887.7+0.9
SpeciEvalSafety98.299.0+0.8
SpeciEval (Refusal-Aware Retries, Historical)Safety98.299.0+0.8
SpeciEval (Pooled English Runs)Safety98.299.0+0.8
InteractiveBench (task pass)Agents30.030.8+0.8
APEX-AgentsProfessional Work38.537.7−0.8
MCP Atlas public (Kimi K3 100-turn evaluation)Agents82.883.6+0.8
MMLU Pro - PsychologyGeneral QA88.589.2+0.8
AA-Omniscience IndexIndexes20.522.0+0.7
LLM Chess EloGames15321550+0.7
MMLU Pro - EconomicsGeneral QA89.990.6+0.7
MMLU Pro - ChemistryGeneral QA91.392.0+0.7
Bench to the Future 3 — Pooled ScoreForecasting0.10.1+0.7
ERQAMultimodal70.571.2+0.7
SEA-HELMGeneral QA76.377.0+0.7
LMArena Text Arena EloChat & Writing14811485+0.6
GBA Eval OverallGames53.252.6−0.6
Finance Agent v2 - General Qualitative AnalysisFinance & Accounting65.664.9−0.6
Long-Horizon Terminal-Bench 1.0 (mean reward)Code44.545.1+0.6
MMLU Pro - PhilosophyGeneral QA87.888.4+0.6
Terminal-Bench 2.1 Extended (Mercor held-out)Code38.037.4−0.6
PubMedQA (SecureBio)Health & Medicine80.581.1+0.6
SWE Atlas - Codebase QnACode45.446.0+0.6
MMLU Pro - BiologyGeneral QA94.094.6+0.6
GPQA DiamondGeneral QA95.594.9−0.6
MMLU Pro - MathMath94.795.3+0.5
CharXiv ReasoningMultimodal84.184.6+0.5
HealthBench (length-adjusted)Health & Medicine56.557.0+0.5
MMVUMultimodal81.781.2−0.5
LegalBenchLegal86.587.0+0.5
LLM Creative Story WritingChat & Writing2.42.5+0.4
HealthBench Consensus (updated implementation)Health & Medicine95.795.3−0.4
AA-LCR v1.1Search84.384.0−0.3
DTBenchGeneral QA96.095.7−0.3
MMLU Pro - HistoryGeneral QA81.681.9+0.3
MMLU Pro - HealthGeneral QA84.484.1−0.2
TxBench Antibody Discovery V0Science33.633.8+0.2
TaxEval v2Finance & Accounting75.074.8−0.2
OpenAI MRCR v2 8-needle - 512K-1MGeneral QA74.073.8−0.2
Video-MME (with subtitles)Multimodal89.389.5+0.2
BioSecBench-Surveillance V1 Clean RerunsScience47.147.3+0.2
LMArena Agent Arena SteerabilityAgents6.96.5−0.2
BabaIsBench Stage 0: The IntroGames100.0100.00.0
CRM-Agent-BenchAgents81.381.30.0
FrontierMath ErdősMath0.00.00.0
HardGeoBench <=100miMultimodal40.040.00.0
MCPMark-VerifiedAgents92.992.90.0
OTIS Mock AIME 2024-2025Math100.0100.00.0
Ramp SWE-Bench (78 published tasks)Code83.383.30.0
Visual Acuity 4K Threshold RowMultimodal12.012.00.0
WordleBenchGames97.097.00.0

GPT-5.5vsGPT-5.6 Solnormalised score marginsData Nov 15, 2026Data: Vals AI, Artificial Analysis, benchmarks.bio and 98 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…