Atlas

Compare Models

Claude Fable 5.1 vs GPT-5.6 Terra

Claude Fable 5.1 leads by 5.0 capability points (≈95% interval +2.8 to +7.3): a clear lead over GPT-5.6 Terra · P(higher) ≈ 100%

Wins 138 of 166 shared benchmarks (27 lost, 1 tied)

Capability vs. relative cost
Claude Fable 5.1GPT-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.

Claude Fable 5.1vsGPT-5.6 Terracapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins138 – 27 · 166 shared, 1 tied

Comparing matching setups only (54 benchmarks excluded) · suites counted once
← Claude Fable 5.1GPT-5.6 Terra →

Claude Fable 5.1vsGPT-5.6 Terrawins by categoryData Nov 15, 2026Data: Mercor, Vals AI, Center for AI Safety and 47 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 Claude Fable 5.1.

VCT-Refusal (CAIS)Safety98.910.7−88.2
HPCT Refusal (CAIS)Safety100.025.0−75.0
LMArena Agent Arena Net ImprovementAgents14.10.6−44.3
Document ArenaProfessional Work15151468−42.2
GameBench 2Games94.053.5−40.6
Medical Long Context Reasoning (MLCR-AA)Health & Medicine71.131.7−39.4
LisanBench Path LengthGames151003303−36.7
Gray Swan IPI Attack Success (k=15)Safety1.037.3−36.3
LisanBench Difficulty-WeightedGames45731197−36.2
HardGeoBench Avg DistanceMultimodal8451991−35.3
BioTIER RefusalSafety99.565.6−33.9
APEX-SWECode63.631.1−32.5
Terminal-Bench 4.0Code57.926.3−31.6
LMArena Agent Arena Net Improvement (Steering Burden)Agents14.8-1.8−31.5
SM BenchSafety89.059.0−30.0
ProofBench v1.1Math100.074.0−26.0
ALE-BenchCode28492154−24.6
Terminal-Bench Science 0.1 (Vals)Science34.310.0−24.3
BullshitBench v2 LeaderboardSafety77.054.0−23.0
Mystery Game PuzzlesGames58.035.0−23.0
SWE Atlas Codebase QnA Extended (Mercor held-out)Code38.016.0−22.0
MindCube TinyMultimodal84.562.6−21.9
GDPval Extended (Mercor held-out rubric score)Professional Work68.747.2−21.5
NoCoT-Bench NCRI 15.2General QA12889.0−21.4
HackerRank-BenchCode47.526.1−21.4
EnigmaEvalGeneral QA44.223.1−21.1
Last Translation Benchmark (LTBv1-eval, blind)Chat & Writing32.912.2−20.7
APEX-AccountingFinance & Accounting61.040.3−20.7
Agent Red Teaming k=100 (CAIS)Safety34.254.6−20.3
HardGeoBench <=10miMultimodal45.025.0−20.0
Toloka Arena (v1)Agents60.741.4−19.3
AA-Briefcase EloAgents16621330−19.3
SWE-bench VerifiedCode96.677.8−18.8
Olam Social Poker EloGames14941511+18.5
DAYJOB: FinanceFinance & Accounting19.81.8−18.0
GeneBench-Pro Public Release SubsetScience26.08.0−18.0
HardGeoBench <=100miMultimodal53.836.3−17.5
SWE-bench Verified Extended (Mercor held-out)Code61.378.7+17.4
FrontierMath Tier 4 (v2) - PrivateMath87.870.7−17.1
NoCoT-Bench NCKI (kspine v3)General QA128104−17.1
Terminal-Bench 2.1 Extended (Mercor held-out)Code19.536.4+16.9
LMArena Agent Arena Praise vs ComplaintAgents31.6-1.9−16.8
TextQuests Average Cumulative HarmSafety27.122.2+16.5
TextQuests Game ProgressGames56.640.5−16.1
Shroombench (mushroom identification, top-1)Multimodal51.035.0−16.0
Furniture Assembly BenchmarkMultimodal70.054.2−15.8
Design Arena Overall Code CategoriesCode13311238−15.4
τ³-Banking (Mercor run, bm25_grep retrieval)Agents54.038.8−15.2
Creative Writing v3 EloChat & Writing21551856−15.1
Harvey LAB (Mercor run, undisclosed ~213-task cohort)Legal17.42.3−15.1
MedXpertQA MM Extended (Mercor held-out)Health & Medicine66.051.0−15.0
WeirdML (v2)Science92.978.3−14.6
BenchCAD - geometric overlapMultimodal92.678.2−14.4
HardGeoBench Country AccuracyMultimodal58.845.0−13.7
GDPval-AA v2.1Professional Work17651461−13.4
GDPval-AA v2Professional Work17451479−13.1
Hedge-BenchFinance & Accounting26.813.8−13.0
LiveBench 2.0Agents88.075.0−13.0
PCT Political Inconsistency (CAIS)Safety33.446.2−12.8
Conceptual Reasoning IndexIndexes74.965.7−12.7
SpatialViz-BenchMultimodal77.965.2−12.7
Tax Agent BenchFinance & Accounting77.965.2−12.7
ChartographyMultimodal46.234.0−12.2
BabaIsBench Stage 0: The IntroGames100.088.0−12.0
CVDP (517-problem set)Code62.650.9−11.7
Extended NYT Connections 940 (Quadratic v1)Games90.078.4−11.6
Artificial Analysis Intelligence Index v4.3Indexes53.442.3−11.1
Blueprint-Bench 2Multimodal0.40.3−11.1
SpeciEval (Refusal-Aware Retries, Historical)Safety87.298.1+10.8
SpeciEval (Pooled English Runs)Safety87.298.1+10.8
LMArena Agent Arena Confirmed SuccessAgents14.4-7.2−10.8
Humanity's Last ExamGeneral QA46.535.9−10.6
LMCAGeneral QA65.555.0−10.5
CursorBench 4.0Code51.841.3−10.5
Excel Modeling BenchmarkFinance & Accounting76.766.2−10.5
ProgramBench (Mercor 100-task subset)Code11.31.0−10.3
ReactBench (39 tasks)Code49.239.0−10.3
ChessBench EloGames1095925−10.1
MedCodeHealth & Medicine53.543.4−10.1
SUPERChemScience74.764.9−9.8
BuseyBench SVGMultimodal7.06.1−9.5
Finance Agent v2 - General Qualitative AnalysisFinance & Accounting77.067.8−9.2
ACCoRDGeneral QA83.774.7−9.0
Finance Agent v2 - Market AnalysisFinance & Accounting69.560.6−8.9
Roboflow Vision Evals (October 2026, five tasks)Multimodal74.766.0−8.7
AA-LCR v1.1Search87.579.0−8.5
MedScribeHealth & Medicine91.382.9−8.4
SciCode (main problems)Science24.215.8−8.4
SciCode 1.0.1 (Artificial Analysis)Code63.155.0−8.1
CadQueryEval (Corrected Geometry)Code76.068.0−8.0
Concept Synth INDUCTION Challenge100 Round 1General QA33.025.0−8.0
DAYJOB: HealthcareHealth & Medicine9.61.6−8.0
MedXpertQA MM (200-question set)Health & Medicine83.775.7−8.0
ChessBench CoherenceGames61.969.7+7.8
DeepResearch Bench II (Mercor 130-task set)Search55.047.2−7.8
LMArena Agent Arena Steering BurdenAgents14.8-0.5−7.7
AutomationBench 1.0.6 (Zapier)Agents22.414.8−7.6
BrowseComp (Mercor 130-question subset)Search78.385.8+7.5
Roboflow Vision Evals (September 2026 protocol)Multimodal81.373.8−7.5
MASKSafety85.993.3+7.4
Finance Agent v2 - PrecedentsFinance & Accounting32.224.8−7.4
Senior SWE-BenchCode34.727.4−7.3
Longform Creative WritingChat & Writing85.378.0−7.3
DeepSWE v1.1Code67.360.2−7.1
Finance Agent v2 - General Quantitative AnalysisFinance & Accounting79.772.7−7.1
ERQAMultimodal70.363.6−6.7
MMMU ProMultimodal86.179.5−6.6
IntPhys 2Multimodal67.761.3−6.4
Terminal-Bench 2.1Code81.688.0+6.4
FrontierCode 2.0 (Artificial Analysis)Code30.323.9−6.4
Machiavelli Deception (CAIS)Safety82.480.4+6.2
AdvancedIFChat & Writing84.278.1−6.1
ARC-AGI-2General QA90.083.9−6.1
LMArena Text Arena EloChat & Writing15031468−6.1
GBENCH One-shot Coding GScoreGames69.363.3−6.0
GDPval (Mercor 213-task rubric score)Professional Work86.680.7−5.9
Vision Arena OverallMultimodal12891266−5.9
LMArena Agent Arena Bash RecoveryAgents11.60.9−5.4
APEX-Agents 1.1Professional Work77.171.8−5.3
Finance Agent v2 - All PassFinance & Accounting47.842.8−5.0
Epoch Capabilities Index (ECI)Indexes167161−5.0
MMLU ProGeneral QA93.088.1−4.9
DistressBenchSafety71.576.4+4.9
Financial Coaching BenchFinance & Accounting75.771.0−4.7
Finance Agent v2 - AdjustmentsFinance & Accounting54.449.7−4.7
LMArena Agent Arena Bash ReliabilityAgents10.10.7−4.7
LiveCodeBench (Vals AI public dataset)Code90.585.9−4.6
APEX-1 (Mercor, mean score)Professional Work73.869.5−4.3
DTBenchGeneral QA97.693.3−4.3
FrontierMath Tiers 1-3 (v2) - PrivateMath90.286.0−4.2
GPQA DiamondGeneral QA93.989.9−4.0
Eval Connections One-Shot with TrapsGames82.086.0+4.0
SAB RankingForecasting-0.1-0.0+3.5
SAB ResearchFinance & Accounting96.893.4−3.5
MortgageTax (Vals all 1,258 documents)Finance & Accounting70.867.3−3.5
LegalBenchLegal88.585.1−3.4
Finance Agent v2 - Disclosure AnalysisFinance & Accounting70.366.9−3.3
SWE-bench MultilingualCode91.194.4+3.3
CritPt v2 (agentic, avg@5)Science72.969.7−3.1
Concept Synth INDUCTION Challenge64 Round 1General QA34.437.5+3.1
Finance Agent v2 - Financial ModelingFinance & Accounting25.323.0−2.3
LMArena Agent Arena SteerabilityAgents8.44.1−2.1
BioSecBench-RespiratoryVirus-Surveillance V1 (110 evaluations)Science3.65.8+2.1
Ruby LLM Benchmarks — Program FixerCode76.074.0−2.0
SWE Atlas - Codebase QnACode43.345.2+1.9
Creative Writing v3 Rubric ScoreChat & Writing16.315.9−1.8
Finance Agent v2 - Earnings AnalysisFinance & Accounting75.577.3+1.8
SAGEMath48.547.0−1.5
ITBench-AAAgents49.551.0+1.5
AA-LCR Extended (Mercor held-out)Search73.074.4+1.4
ObviousBenchGeneral QA99.397.9−1.4
CritPt (avg@5)Science31.130.0−1.1
CharXiv Extended (Mercor held-out)Multimodal82.281.1−1.1
Finance Agent v2 - ComparablesFinance & Accounting46.047.0+1.0
ARC-AGI-1 Semi-PrivateGeneral QA97.596.5−1.0
CharXiv (descriptive and reasoning combined)Multimodal95.194.1−1.0
Chess Puzzles (task version 2.0)Games48.047.0−1.0
RiemannBenchMath65.666.4+0.8
ChessBench AccuracyGames71.371.9+0.6
HealthBench Professional (Length-Adjusted)Health & Medicine58.157.7−0.4
OTIS Mock AIME 2024-2025Math100.099.7−0.3
TaxEval v2Finance & Accounting76.076.2+0.2
MMMU-Pro Extended (Mercor held-out)Multimodal47.647.4−0.2
HieroglyphBenchMultimodal20.420.5+0.1
LMArena Agent Arena Tool HallucinationAgents0.81.0+0.1
AIME 2025Math100.0100.00.0

Claude Fable 5.1vsGPT-5.6 Terranormalised score marginsData Nov 15, 2026Data: Mercor, Vals AI, Center for AI Safety 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…