Atlas

Compare Models

Claude Sonnet 5.5 vs GPT-5.6 Sol

Claude Sonnet 5.5 leads by 3.8 capability points (≈95% interval +1.3 to +6.2): a clear lead over GPT-5.6 Sol · P(higher) ≈ 100%

Wins 110 of 155 shared benchmarks (42 lost, 3 tied)

Capability vs. relative cost
Claude Sonnet 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.

Claude Sonnet 5.5vsGPT-5.6 Solcapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins110 – 42 · 155 shared, 3 tied

Comparing matching setups only (64 benchmarks excluded) · suites counted once
← Claude Sonnet 5.5GPT-5.6 Sol →

Claude Sonnet 5.5vsGPT-5.6 Solwins by categoryData Nov 15, 2026Data: Artificial Analysis, benchmarks.bio, Vals AI and 54 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 Claude Sonnet 5.5.

KernelBench Mega Normalized (RTX PRO 6000)Code90.67.4−83.2
Medical Long Context Reasoning (MLCR-AA)Health & Medicine75.026.1−48.9
YC-BenchGames1741315728001−35.4
Terminal-Bench-Science 0.1Science53.322.4−31.0
FrontierSWE V2Code61.932.2−29.7
MetagenomicsBench V0Science60.031.3−28.7
MetagenomicsBench V1Science60.031.3−28.7
LisanBench Path LengthGames137874699−28.2
Terminal-Bench Science 0.1 (Vals)Science47.120.0−27.1
EnterpriseBench: CoreCraftProfessional Work70.846.2−24.6
Terminal-Bench 4.0Code61.837.3−24.5
ChessBench CoherenceGames64.689.1+24.5
SM BenchSafety85.461.1−24.3
Terminal-Bench 4.0 (Artificial Analysis)Code63.639.9−23.7
Humanity's Last Exam DiamondGeneral QA53.330.4−22.9
ChessBench EloGames8571218+21.5
LMArena Agent Arena Net Improvement (Steering Burden)Agents16.04.6−21.4
HardGeoBench Avg DistanceMultimodal6891381−21.4
AA-Briefcase v1.1 EloAgents18161479−18.5
DAYJOB: FinanceFinance & Accounting24.05.5−18.5
APEX-SWECode66.448.0−18.4
Furniture Assembly BenchmarkMultimodal75.056.7−18.3
TxBench-Antibody-Development V1 (60 evaluations)Science36.718.9−17.8
ActiveVisionMultimodal41.223.5−17.7
TxBench Antibody Discovery V1 (100 evaluations)Science54.737.3−17.3
ARC-AGI-3 (Semi-Private)General QA24.37.0−17.3
ProofBench v1.1Math100.083.0−17.0
LisanBench Difficulty-WeightedGames29831403−16.9
Code MigrationCode69.852.9−16.9
ChartographyMultimodal61.645.0−16.6
VariantBench V1 Clean RerunsScience51.735.3−16.4
BioSecBench-Function V1 Clean RerunsScience51.735.5−16.2
MysteryMechanismScience49.133.3−15.8
GeneBench-Pro Public Release SubsetScience39.024.0−15.0
HardGeoBench Country AccuracyMultimodal62.547.5−15.0
TxBench Oligo Discovery V1 (113 evaluations)Science51.937.2−14.7
LLM Chess EloGames11951550+14.6
VIALSMultimodal41.026.5−14.5
CursorBench 4.0Code55.541.7−13.8
HardGeoBench <=100miMultimodal53.840.0−13.7
DAYJOB: HealthcareHealth & Medicine15.21.6−13.6
Extended NYT Connections 940 (Quadratic v1)Games80.593.8+13.3
Nimble Web ContextSearch67.280.0+12.8
Humanity's Last ExamGeneral QA56.944.5−12.4
LMArena Agent Arena Praise vs ComplaintAgents30.56.3−12.1
BullshitBench v2 LeaderboardSafety60.048.0−12.0
TaxCalcBench TY25Finance & Accounting18.030.0+12.0
Vibe Code Bench v1.1Code92.480.5−11.9
AutomationBench-AA 1.0.6Agents71.860.1−11.7
BioSecBench-RespiratoryVirus-Surveillance V1 (110 evaluations)Science17.35.8−11.5
Conceptual Reasoning IndexIndexes77.168.7−11.5
GameBench 2Games91.480.1−11.4
OncologyBench Strict (25 tasks, October 9 rubric)Health & Medicine29.318.7−10.7
HealthBenchHealth & Medicine65.955.3−10.6
AA-AnalystAgent (pass^5)Professional Work57.547.5−10.0
RedlineBenchLegal40.350.3+10.0
LMCAGeneral QA68.358.5−9.8
ACCoRDGeneral QA85.775.9−9.8
GDPval-AA v2.1Professional Work18381619−9.7
NoCoT-Bench NCKI (kspine v3)General QA107121+9.4
Vending-Bench 2Games111659619−9.4
Artificial Analysis Intelligence Index v4.3.2Indexes56.047.0−9.0
MedCodeHealth & Medicine52.944.0−8.9
Olam Social Poker EloGames15261534+8.7
ARC-AGI-2General QA83.992.5+8.6
LMArena Agent Arena Confirmed SuccessAgents19.52.3−8.6
IOI (Vals v2)Code83.191.2+8.1
Terminal-Bench 2.1 Extended (Mercor held-out)Code45.537.4−8.1
SpeciEval (Refusal-Aware Retries, Historical)Safety91.099.0+8.1
SpeciEval (Pooled English Runs)Safety91.099.0+8.1
NoCoT-Bench NCRI 15.2General QA95.1109+7.9
Concept Synth INDUCTION Challenge64 Round 1General QA50.057.8+7.8
LMArena Agent Arena Bash RecoveryAgents15.00.7−7.2
Mystery Game PuzzlesGames65.058.0−7.0
Eval Connections One-Shot with TrapsGames76.083.0+7.0
Artificial Analysis Progress Index v1Indexes351331−6.7
Vision Arena OverallMultimodal12631288+6.5
RiemannBench (November 2026 rescoring)Math76.883.2+6.4
LAB-Bench ProtocolQA (SecureBio)Science88.381.9−6.4
HardGeoBench <=10miMultimodal36.330.0−6.2
LiveBench 2.0Agents82.576.3−6.2
MedScribeHealth & Medicine91.185.2−5.9
Blueprint-Bench 2Multimodal0.40.3−5.7
Short Form Bench v1.0Games23552410+5.5
CVDP (517-problem set)Code60.655.4−5.2
Roboflow Vision Evals (October 2026, five tasks)Multimodal77.672.4−5.2
AA-Omniscience IndexIndexes32.322.0−5.2
ProgramBenchCode6.51.5−5.0
Roboflow Vision Evals (September 2026 protocol)Multimodal83.879.0−4.8
SpatialBench V1 Clean RerunsScience72.868.1−4.6
SWE-TogetherCode62.457.8−4.6
LMArena Agent Arena Steering BurdenAgents14.55.5−4.5
EpiBench V1 Clean RerunsScience30.225.8−4.4
AA-LCR v1.1Search79.784.0+4.3
Bio Capabilities Index (SecureBio)Indexes175169−4.3
BuseyBench SVGMultimodal6.56.1−4.2
BrowseComp (Mercor 130-question subset)Search86.690.6+4.0
FrontierCode 1.1 Extended - ScoreCode64.460.6−3.8
ObviousBenchGeneral QA100.096.5−3.5
Excel Modeling BenchmarkFinance & Accounting75.772.3−3.4
SciCode 1.0.1 (Artificial Analysis)Code61.057.8−3.2
SWE Atlas - Codebase QnACode50.353.5+3.2
MMMU ProMultimodal85.882.7−3.1
Toloka Arena (v3)Agents64.061.0−3.0
TxBench Preclinical Pharmacology V1 Clean RerunsScience64.361.3−3.0
LAB-Bench LitQA2 (SecureBio)Science55.852.8−3.0
FrontierMath ErdősMath2.90.0−2.9
WMDP-Bio (SecureBio)Science92.689.7−2.9
CritPt (avg@5)Science31.428.6−2.9
VCT v2 (SecureBio)Science52.249.4−2.8
BigFinanceBench private holdout (final-answer accuracy)Finance & Accounting41.544.2+2.7
Epoch Capabilities Index (ECI)Indexes166163−2.7
SpatialBench-Long V1 Clean RerunsScience34.937.5+2.6
FrontierMath Tier 4 (v2) - PrivateMath80.582.9+2.4
MBCT (SecureBio)Science58.861.1+2.3
GPQA-Bio (SecureBio)Science87.685.3−2.3
MaxxBenchCode66.969.1+2.2
Terminal-Bench 2.1Code87.389.5+2.2
DeepsecBench — Revalidated V1 (233 issues)Code37.535.3−2.2
DeepSWE v1.1Code70.572.7+2.2
LMArena Text Arena EloChat & Writing14721485+2.2
WCB (SecureBio)Science66.564.4−2.1
Veracode GenAI Code Security (Fall 2026)Cybersecurity72.070.0−2.0
LAB-Bench CloningScenarios (SecureBio)Science97.098.8+1.8
VCT (SecureBio)Science52.554.3+1.8
DTBenchGeneral QA97.395.7−1.6
CritPt v2 (agentic, avg@5)Science75.777.1+1.4
ALE-BenchCode24482410−1.4
Ruby LLM Benchmarks — Program FixerCode73.875.2+1.3
BioSecBench-Structure V1 (58 evaluations)Science21.820.7−1.1
ARC-AGI-1 Semi-PrivateGeneral QA96.597.5+1.0
Concept Synth INDUCTION Challenge100 Round 1General QA44.043.0−1.0
DistressBenchSafety74.673.6−1.0
HPCT (SecureBio)Science77.276.3−0.9
SAGEMath51.852.6+0.8
ScBench-Long V1 Clean RerunsScience30.329.6−0.8
GDP.pdfProfessional Work31.430.7−0.7
Public Benefits Bench v1.1Professional Work67.266.5−0.7
LMArena Agent Arena Bash ReliabilityAgents12.711.4−0.7
ChessBench AccuracyGames75.876.4+0.6
ScBench V1 Clean RerunsScience60.359.8−0.5
Prophet Arena Arena Score (Fixed Context)Forecasting85.384.8−0.4
Harvey's Legal Agent BenchmarkLegal2.92.5−0.4
FrontierMath Tiers 1-3 (v2) - PrivateMath88.889.1+0.4
ForecastBench Preliminary Dataset Brier IndexForecasting61.161.4+0.3
GBENCH One-shot Coding GScoreGames67.467.7+0.3
SWE-Bench Pro V2 — FullCode95.895.5−0.3
DeepResearch Bench II (Mercor 130-task set)Search51.751.5−0.2
PubMedQA (SecureBio)Health & Medicine81.381.1−0.2
MMLU-Bio (SecureBio)Science91.091.1+0.1
LMArena Agent Arena Tool HallucinationAgents0.91.0+0.1
Financial Coaching BenchFinance & Accounting74.974.9−0.1
Chess Puzzles (task version 2.0)Games59.059.00.0
OTIS Mock AIME 2024-2025Math100.0100.00.0
WellnessBenchSafety57.057.00.0

Claude Sonnet 5.5vsGPT-5.6 Solnormalised score marginsData Nov 15, 2026Data: Artificial Analysis, benchmarks.bio, Vals AI and 54 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…