Atlas

Compare Models

GPT-5.5 vs GPT-5.6 Terra

GPT-5.6 Terra leads by 0.7 capability points (≈95% interval −1.3 to +2.7): statistically tied · P(higher) ≈ 75%

GPT-5.5 wins 150 of 287 shared benchmarks (112 lost, 25 tied)

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

Benchmark wins150 – 112 · 287 shared, 25 tied

Comparing matching setups only (18 benchmarks excluded) · suites counted once
← GPT-5.5GPT-5.6 Terra →

GPT-5.5vsGPT-5.6 Terrawins by categoryData Nov 15, 2026Data: Vals AI, Mercor, Artificial Analysis and 70 othersVillage LLM Atlasatlas.village.dev

Margins

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

SnakeBench Win RateGames90.050.0−40.0
SWE-bench Verified Extended (Mercor held-out)Code39.378.7+39.4
ChessBench CoherenceGames100.069.7−30.3
GraphWalks BFS - 1MGeneral QA45.471.2+25.8
ChessBench EloGames1349925−25.2
LisanBench Difficulty-WeightedGames33161197−22.7
Mystery Game PuzzlesGames56.035.0−21.0
ProofBenchMath50.071.0+21.0
SnakeBench RatingGames37.429.8−20.3
LisanBench Path LengthGames98453303−20.3
PMPP-HardCode52.231.9−20.3
Harvey's Legal Agent Benchmark - Data Privacy CybersecurityLegal20.00.0−20.0
Harvey's Legal Agent Benchmark - Energy Natural ResourcesLegal20.00.0−20.0
Harvey's Legal Agent Benchmark - Structured Finance SecuritizationLegal0.020.0+20.0
SimpleQA VerifiedGeneral QA63.143.2−19.9
YC-BenchGames1206045636144−19.9
EQ-Bench 4Chat & Writing12951220−19.7
SnakeBench Average ApplesGames16.512.7−19.0
MindCube TinyMultimodal79.962.6−17.3
SWE-bench MultilingualCode77.794.4+16.7
Chess Puzzles (task version 2.0)Games63.047.0−16.0
Aikido Known-CVE Detection (Pass@3 Union)Cybersecurity73.188.5+15.4
Document ArenaProfessional Work14841468−14.9
SnakeBench Best ApplesGames26.021.0−14.7
EnigmaEvalGeneral QA37.223.1−14.1
Medical Long Context Reasoning (MLCR-AA)Health & Medicine17.831.7+13.9
τ²-bench (Telecom)Agents93.980.4−13.5
Agent Red Teaming k=100 (CAIS)Safety41.554.6−13.0
SWE-bench VerifiedCode82.695.4+12.8
Extended NYT Connections 940-Puzzle SetGames97.584.7−12.8
PointerbenchMultimodal59.947.3−12.6
ScBench-Long V0 VerifiedScience22.734.9+12.1
BabaIsBench Stage 0: The IntroGames100.088.0−12.0
LMArena Agent Arena Net ImprovementAgents4.20.6−11.9
AA-Briefcase EloAgents11361337+11.7
ChessBench — Historical Output RepairGames29.340.7+11.3
RiemannBenchMath55.266.4+11.2
Toloka Arena (v1)Agents52.341.4−10.9
Olam Social Poker EloGames15211511−10.9
DeepSWE v1.1Code64.453.8−10.6
AA-Omniscience IndexIndexes20.50.1−10.2
Furniture Assembly BenchmarkMultimodal44.254.2+10.0
Arena Code WebDev — FullstackCode15271565+10.0
LLM Chess EloGames15321293−9.8
Terminal-Bench 2.1Code78.288.0+9.8
MedXpertQA MM Extended (Mercor held-out)Health & Medicine60.851.0−9.8
VariantBenchScience29.738.7+9.0
SpatialViz-BenchMultimodal74.265.2−9.0
A-FantasiaGeneral QA34.743.3−8.7
A-Fantasia (Valid Attempts with Retries)Games34.743.3−8.7
TextQuests Game ProgressGames48.940.5−8.4
AHK-EvalCode97.288.9−8.3
EnterpriseOps-Gym-AAAgents46.638.5−8.1
OpenAI MRCR v2 8-needle - 256K-512KGeneral QA81.589.6+8.1
Finance Agent v2 - Disclosure AnalysisFinance & Accounting58.966.9+8.0
CadQueryEval (Corrected Geometry)Code76.068.0−8.0
CadQueryEvalCode64.056.0−8.0
ScBench-LongScience15.923.8+7.9
Humanity's Last ExamGeneral QA43.835.9−7.9
APEX-AccountingFinance & Accounting48.240.3−7.9
BioTIER RefusalSafety73.465.6−7.8
ChessBench AccuracyGames78.871.9−7.4
GDPval Extended (Mercor held-out rubric score)Professional Work54.547.2−7.3
BullshitBench v2 LeaderboardSafety47.054.0+7.0
ERQAMultimodal70.563.6−6.9
GeneBench-ProScience12.018.8+6.8
WeirdML (v2)Science84.978.3−6.6
CursorBench 3.2Code58.464.9+6.5
GameBench 2Games59.953.5−6.5
SM BenchSafety65.459.0−6.4
Ramp SWE-Bench (78 published tasks)Code83.376.9−6.4
BrowseComp Extended (Mercor held-out)Search19.025.3+6.3
LiveBench 2.0Agents68.775.0+6.3
PCT Political Inconsistency (CAIS)Safety40.046.2−6.2
NoCoT-Bench NCRI 15.2General QA10089.0−6.2
Opus Magnum CampaignGames53.947.7−6.1
Frontier Financial JudgementFinance & Accounting52.446.3−6.1
WordleBenchGames97.091.0−6.0
ACCoRDGeneral QA80.574.7−5.8
SRE Bench — Fully SolvedCode3.89.5+5.7
MedCodeHealth & Medicine49.143.4−5.7
MRCR v2 (8-needle, Context Arena) - 128K pointwiseGeneral QA85.880.2−5.7
APEX-SWECode36.631.1−5.5
Blueprint-Bench 2Multimodal0.40.3−5.4
Diligence Stack Agent BenchFinance & Accounting82.076.7−5.3
AA-LCR v1.1Search84.379.0−5.3
VCT-Refusal (CAIS)Safety16.010.7−5.3
ITBench-AAAgents45.851.0+5.2
SpatialBench V0 VerifiedScience65.270.4+5.2
HealthBench ProfessionalHealth & Medicine57.262.4+5.2
VCT (SecureBio)Science52.347.2−5.1
EpiBenchScience45.039.9−5.0
VulcanBench Frontier v4Code78.583.5+5.0
NoCoT-Bench NCKI (kspine v3)General QA111104−4.9
LAB-Bench CloningScenarios (SecureBio)Science97.392.5−4.8
Vibe Code Bench v1.1Code69.874.6+4.7
Tax Agent BenchFinance & Accounting60.565.2+4.7
IFBench (Artificial Analysis avg@5)Chat & Writing75.971.2−4.6
CritPt (avg@5)Science25.430.0+4.6
SpatialBench-Long V0 VerifiedScience33.337.9+4.6
Terminal-Bench 4.0 (Artificial Analysis)Code14.610.1−4.5
SAGEMath51.547.0−4.5
FrontierCode 1.1 Extended - Pass RateCode62.858.3−4.4
Surface Evolver BenchAgents88.183.8−4.4
GDPval-AA v2.1Professional Work13571456+4.4
Vision Arena OverallMultimodal12831266−4.3
LMArena Agent Arena Bash RecoveryAgents9.50.9−4.3
Arena Code WebDev — Data & AnalyticsCode14671496+4.2
GDPval-AA v2 (normalized)Professional Work49.553.7+4.1
Finance Agent v2 - Earnings AnalysisFinance & Accounting73.277.3+4.1
SnakeBench Total ApplesGames181127−4.1
FrontierCode 1.1 Extended - ScoreCode56.752.6−4.0
Machiavelli Deception (CAIS)Safety81.780.4+4.0
HPCT Refusal (CAIS)Safety29.025.0−4.0
MedScribeHealth & Medicine86.982.9−4.0
Finance Agent v2 - AdjustmentsFinance & Accounting45.749.7+4.0
AA-LCR Extended (Mercor held-out)Search70.574.4+3.9
GDPval (Mercor 213-task rubric score)Professional Work84.680.7−3.9
NanoGPT Speedrun Frontier (human record gap closed)Agents8.111.0+3.9
CorpFin v2 - Max Fitting ContextFinance & Accounting70.066.3−3.7
Artificial Analysis Intelligence Index v4.3.2Indexes38.442.1+3.7
OpenAI Capture-the-Flag challenges - internal release scoreCybersecurity88.191.8+3.7
RedlineBenchLegal44.147.8+3.7
Artificial Analysis Intelligence Index v4.3Indexes38.642.3+3.6
Ruby LLM Benchmarks — Program FixerCode77.674.0−3.6
Arena Code WebDev — Consumer ProductCode14761505+3.6
MMMU-Pro Extended (Mercor held-out)Multimodal43.847.4+3.6
GPQA DiamondGeneral QA93.489.9−3.5
SciCode (main problems)Science12.315.8+3.5
VCT v2 (SecureBio)Science45.341.9−3.4
Arena Code WebDev — Brand & MarketingCode15071530+3.4
Finance Agent v2 - ComparablesFinance & Accounting43.747.0+3.4
SkillsBench with SkillsAgents62.258.9−3.3
AdvancedIFChat & Writing81.378.1−3.2
GraphWalks BFS - 256KGeneral QA73.776.9+3.2
MASKSafety90.193.3+3.2
Finance Agent v2 - All PassFinance & Accounting39.642.8+3.2
Design Arena Overall Code CategoriesCode12571238−3.2
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
RuneBenchGames5.75.9+3.1
BrowseCompSearch84.487.5+3.1
Vals Multimodal IndexIndexes68.165.1−3.0
BioSecBench-Function V0Science36.533.5−3.0
ChartographyMultimodal31.034.0+3.0
ZeroBench Main Mean pass@1 (five samples)Multimodal17.214.2−3.0
Finance Agent v2 - PrecedentsFinance & Accounting21.924.8+3.0
TextQuests Average Cumulative HarmSafety21.422.2−2.9
Vals IndexIndexes68.065.1−2.8
BioSecBench-RefusalSafety37.134.3−2.8
Arena Code WebDev — HTMLCode15431523−2.8
OSWorld 2.0 500-step Partial ScoreAgents47.550.2+2.7
DTBenchGeneral QA96.093.3−2.7
Code MigrationCode45.247.8+2.6
TxBench Antibody Discovery V0Science33.631.0−2.6
Artificial Analysis Agentic IndexIndexes44.947.4+2.5
ToolathlonAgents55.653.1−2.5
CorpFin v2 - Exact PagesFinance & Accounting65.663.2−2.4
Arena Code WebDev — SimulationsCode15471526−2.4
LMArena Text Arena EloChat & Writing14811468−2.3
HieroglyphBenchMultimodal22.820.5−2.3
AutomationBench (Overall)Agents12.915.2+2.3
MRCR v2 (8-needle, Context Arena) - 128K AUCGeneral QA91.789.4−2.3
Terminal-Bench HardCode60.662.9+2.3
Conceptual Reasoning IndexIndexes67.365.7−2.3
Finance Agent v2 - General Qualitative AnalysisFinance & Accounting65.667.8+2.2
Bio Capabilities Index (SecureBio)Indexes158155−2.2
Arena Code WebDev — Reference-Based DesignCode15331516−2.2
Vals Index v2Indexes57.459.6+2.2
GDPval-AA v2Professional Work13721415+2.2
GPQA-Bio (SecureBio)Science87.485.3−2.1
Senior SWE-BenchCode29.527.4−2.1
ObviousBenchGeneral QA100.097.9−2.1
DeepsecBenchCode21.219.2−2.0
DeepsecBench — Revalidated V1Code21.119.1−2.0
DeepsecBench — Revalidated V1 (233 issues)Code21.019.1−2.0
SciCode (Artificial Analysis avg@3)Science55.953.9−2.0
SUPERChemScience66.864.9−1.9
UnityBench-Mini3DCode66.268.0+1.8
AutomationBench 1.0.6 (Zapier)Agents16.614.8−1.8
CharXiv ReasoningMultimodal84.185.9+1.8
DeepResearch Bench II (Mercor 130-task set)Search45.447.2+1.8
FrontierMath Tier 4 (v2) - PrivateMath72.570.7−1.8
Artificial Analysis Coding IndexIndexes74.976.7+1.8
ExploitBench - release-reported exploit progressCybersecurity47.946.2−1.7
Excel Modeling BenchmarkFinance & Accounting64.566.2+1.7
ALE-BenchCode21082154+1.6
CharXiv Extended (Mercor held-out)Multimodal82.781.1−1.6
DiligenceBenchFinance & Accounting47.549.1+1.6
Terminal-Bench 2.1 Extended (Mercor held-out)Code38.036.4−1.6
Agents' Last Exam (ALE)Agents24.526.0+1.5
Arena Code WebDev — Content Creation ToolsCode14941503+1.5
APEX-Agents 1.1Professional Work70.371.8+1.5
ARC-AGI-1 Semi-PrivateGeneral QA95.096.5+1.5
OpenAI MRCR v2 8-needle - 512K-1MGeneral QA74.072.5−1.5
Public Benefits Bench v1.1Professional Work60.962.4+1.5
MortgageTax (Vals all 1,258 documents)Finance & Accounting68.867.3−1.4
LegalBenchLegal86.585.1−1.4
HPCT (SecureBio)Science66.868.2+1.4
IntPhys 2Multimodal59.961.3+1.4
LMArena Agent Arena SteerabilityAgents6.94.1−1.4
Artificial Analysis Progress Index v1Indexes320324+1.4
Voratiq Coding Agent LeaderboardCode18891904+1.4
LAB-Bench LitQA2 (SecureBio)Science49.851.1+1.3
GDP.pdfProfessional Work26.024.7−1.3
SWE Atlas Codebase QnA Extended (Mercor held-out)Code17.316.0−1.3
MRCR v2 (8-needle) - 128K averageGeneral QA94.893.5−1.3
APEX-Agents-AA (avg@3)Professional Work37.738.9+1.3
𝜏³-BankingAgents39.040.2+1.2
HealthBench Hard (length-adjusted)Health & Medicine31.532.7+1.2
MMMU ProMultimodal83.282.0−1.2
TaxEval v2Finance & Accounting75.076.2+1.2
Arena Code WebDev — GamingCode15581548−1.1
GBENCH One-shot Coding GScoreGames62.263.3+1.1
APEX-1 (Mercor, mean score)Professional Work68.469.5+1.1
ARC-AGI-2General QA85.083.9−1.1
Vending-Bench 2Games75247343−1.1
Finance Agent v2 - General Quantitative AnalysisFinance & Accounting73.772.7−1.0
SkillsBench without SkillsAgents49.948.8−1.0
MedXpertQA MM (200-question set)Health & Medicine76.775.7−1.0
ZeroBench Main Questions pass^5Multimodal11.010.0−1.0
Legal Research BenchLegal40.441.3+1.0
Humanity's Last Exam DiamondGeneral QA27.928.8+0.9
LAB-Bench ProtocolQA (SecureBio)Science77.977.0−0.9
SEA-HELMGeneral QA76.375.4−0.9
Arena Code WebDev — ReactCode15051511+0.8
MMLU ProGeneral QA88.988.1−0.8
BioSecBench-SurveillanceScience50.251.0+0.8
BuseyBench SVGMultimodal6.06.1+0.8
Finance Agent v2 (Vals Index subset)Finance & Accounting51.852.6+0.8
Finance Agent v2 (Vals Multimodal Index component)Finance & Accounting51.852.6+0.8
FrontierMath Tiers 1-3 (v2) - PrivateMath85.386.0+0.7
LMCAGeneral QA54.355.0+0.7
AA-LCRSearch73.374.0+0.7
Epoch Capabilities Index (ECI)Indexes160161+0.6
LiveCodeBench (Vals AI public dataset)Code85.385.9+0.6
Creative Writing v3 EloChat & Writing18451856+0.6
Arena Code WebDev — FrontendCode15161521+0.6
CharXiv (descriptive and reasoning combined)Multimodal94.694.1−0.5
CRED v0.92Science96.597.0+0.5
HealthBench (length-adjusted)Health & Medicine56.557.0+0.5
HealthBench Consensus (updated implementation)Health & Medicine95.795.2−0.5
HealthBench Hard (updated implementation)Health & Medicine33.834.3+0.5
CRM-Agent-BenchAgents81.381.6+0.4
Finance Agent v2 - Market AnalysisFinance & Accounting60.360.6+0.3
HealthBench (updated implementation)Health & Medicine58.458.7+0.3
TxBench Oligo Discovery V0Science33.633.3−0.3
OTIS Mock AIME 2024-2025Math100.099.7−0.3
Roboflow Vision EvalsMultimodal76.476.2−0.2
Harvey HLAB — Criteria Pass RateLegal80.079.8−0.2
MBCT (SecureBio)Science57.156.9−0.2
Longform Creative WritingChat & Writing78.278.0−0.2
SpeciEval (Pooled English Runs)Safety98.298.1−0.1
SpeciEvalSafety98.298.1−0.1
SpeciEval (Refusal-Aware Retries, Historical)Safety98.298.1−0.1
PolicyBench US Exact MatchFinance & Accounting83.583.4−0.1
Finance Agent v2 - Financial ModelingFinance & Accounting22.923.0+0.1
WCB (SecureBio)Science59.759.8+0.1
PubMedQA (SecureBio)Health & Medicine80.580.4−0.1
ARC-AGI-3 (Semi-Private)General QA0.40.5+0.1
Creative Writing v3 Rubric ScoreChat & Writing15.915.90.0
AIME 2025Math100.0100.00.0
Harvey's Legal Agent Benchmark - Antitrust CompetitionLegal0.00.00.0
Harvey's Legal Agent Benchmark - Arbitration International Dispute ResolutionLegal0.00.00.0
Harvey's Legal Agent Benchmark - Banking FinanceLegal0.00.00.0
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 - Corporate GovernanceLegal0.00.00.0
Harvey's Legal Agent Benchmark - Corporate M&ALegal0.00.00.0
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 - Funds Asset ManagementLegal0.00.00.0
Harvey's Legal Agent Benchmark - Healthcare Life SciencesLegal0.00.00.0
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 - Litigation Dispute ResolutionLegal0.00.00.0
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
ProgramBenchCode0.50.50.0
SWE Atlas - Codebase QnACode45.245.20.0
Visual Acuity 4K Threshold RowMultimodal12.012.00.0

GPT-5.5vsGPT-5.6 Terranormalised score marginsData Nov 15, 2026Data: Vals AI, Mercor, Artificial Analysis and 70 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…