Atlas

Compare Models

Claude Fable 5 vs GPT-5.6 Terra

Claude Fable 5 leads by 2.8 capability points (≈95% interval +0.5 to +5.0): a clear lead over GPT-5.6 Terra · P(higher) ≈ 99%

Wins 127 of 162 shared benchmarks (31 lost, 4 tied)

Capability vs. relative cost
Claude Fable 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.

Claude Fable 5vsGPT-5.6 Terracapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins127 – 31 · 162 shared, 4 tied

Comparing matching setups only (137 benchmarks excluded) · suites counted once
← Claude Fable 5GPT-5.6 Terra →

Claude Fable 5vsGPT-5.6 Terrawins by categoryData Nov 15, 2026Data: Mercor, Vals AI, Center for AI Safety and 51 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.

VCT-Refusal (CAIS)Safety100.010.7−89.3
HPCT Refusal (CAIS)Safety100.025.0−75.0
YC-BenchGames1977573636144−46.8
Olam Social Poker EloGames15531511−45.7
BenchCAD - geometric overlapMultimodal38.478.2+39.8
LisanBench Difficulty-WeightedGames45621197−36.0
BioTIER RefusalSafety100.065.6−34.4
SimpleBenchGeneral QA81.948.9−33.0
Medical Long Context Reasoning (MLCR-AA)Health & Medicine64.431.7−32.8
LisanBench Path LengthGames136803303−32.2
ExploitBench - release-reported exploit progressCybersecurity78.046.2−31.8
PointerbenchMultimodal78.747.3−31.4
APEX-SWECode58.931.1−27.8
SimpleQA VerifiedGeneral QA70.743.2−27.5
Document ArenaProfessional Work14971468−25.9
LMArena Agent Arena Net ImprovementAgents8.00.6−24.3
Humanity's Last ExamGeneral QA57.835.9−21.9
GameBench 2Games75.253.5−21.8
ProofBench v1.1Math95.074.0−21.0
MindCube TinyMultimodal83.162.6−20.5
Terminal-Bench 2.1 Extended (Mercor held-out)Code16.836.4+19.6
Toloka Arena (v1)Agents60.941.4−19.5
FrontierMath Tier 4 (v2) - PrivateMath90.270.7−19.5
BrowseComp Extended (Mercor held-out)Search44.525.3−19.2
Furniture Assembly BenchmarkMultimodal35.854.2+18.3
Terminal-Bench 4.0Code44.526.3−18.3
GDPval Extended (Mercor held-out rubric score)Professional Work65.347.2−18.1
LMArena Agent Arena Net Improvement (Steering Burden)Agents7.3-1.8−17.3
Vending-Bench 2Games45307343+17.0
Mystery Game PuzzlesGames52.035.0−17.0
NoCoT-Bench NCKI (kspine v3)General QA128104−17.0
Odysseys2Agents61.544.6−16.9
SWE-Bench Pro (Public Dataset)Code80.363.4−16.9
Humanity's Last Exam DiamondGeneral QA45.628.8−16.8
NoCoT-Bench NCRI 15.2General QA11989.0−16.5
APEX-AccountingFinance & Accounting56.440.3−16.1
SWE Atlas Codebase QnA Extended (Mercor held-out)Code0.016.0+16.0
OSWorld 2.0 500-step Partial ScoreAgents66.150.2−15.9
Vibe Code Bench v1.1Code90.474.6−15.8
TextQuests Game ProgressGames56.140.5−15.6
Agent Red Teaming k=100 (CAIS)Safety39.454.6−15.2
WeirdML (v2)Science91.978.3−13.7
AdvancedIFChat & Writing65.278.1+12.9
ChessBench CoherenceGames82.669.7−12.9
MedCodeHealth & Medicine56.143.4−12.7
Opus Magnum CampaignGames60.247.7−12.4
τ²-bench (Telecom)Agents98.586.3−12.3
SWE Atlas - Codebase QnACode33.145.2+12.1
GDPval-AA v2Professional Work17231479−12.0
BabaIsBench Stage 0: The IntroGames100.088.0−12.0
CadQueryEval (Corrected Geometry)Code80.068.0−12.0
HLE Extended (Mercor held-out)General QA58.770.7+12.0
Humanity's Sixth SenseMultimodal34.522.7−11.8
AA-Briefcase EloAgents15301330−11.6
Ramp SWE-Bench (78 published tasks)Code88.576.9−11.5
Surface Evolver BenchAgents95.083.8−11.2
Vision Arena OverallMultimodal13091266−11.0
ALE-BenchCode24632154−10.9
Artificial Analysis Intelligence IndexIndexes62.151.6−10.5
Extended NYT Connections 940-Puzzle SetGames95.184.7−10.4
ACCoRDGeneral QA85.074.7−10.3
MMMU-Pro Extended (Mercor held-out)Multimodal37.647.4+9.8
DeepSWE v1.1Code69.960.2−9.7
Conceptual Reasoning IndexIndexes72.765.7−9.6
SpatialViz-BenchMultimodal73.965.2−8.7
MMLU ProGeneral QA79.488.1+8.7
LiveBench 2.0Agents83.675.0−8.6
ToolathlonAgents61.753.1−8.6
LMArena Agent Arena Praise vs ComplaintAgents14.3-1.9−8.1
GDPval-AA v2 (normalized)Professional Work62.254.1−8.1
Blueprint-Bench 2Multimodal0.40.3−7.8
Excel Modeling BenchmarkFinance & Accounting73.766.2−7.5
Senior SWE-BenchCode34.727.4−7.3
DAYJOB: FinanceFinance & Accounting9.01.8−7.2
ChessBench EloGames1046925−7.2
Terminal-Bench 4.0 (Artificial Analysis)Code42.435.4−7.1
IOICode72.365.3−7.0
MMMU ProMultimodal72.779.5+6.8
CorpFin v2Finance & Accounting71.865.3−6.5
Diligence Stack Agent BenchFinance & Accounting83.276.7−6.5
PCT Political Inconsistency (CAIS)Safety39.846.2−6.4
RiemannBenchMath60.066.4+6.4
LMArena Text Arena EloChat & Writing15041468−6.3
SciCode (Artificial Analysis avg@3)Science60.253.9−6.2
LMCAGeneral QA61.155.0−6.2
LMArena Agent Arena Confirmed SuccessAgents5.1-7.2−6.2
Roboflow Vision EvalsMultimodal82.276.2−6.0
GeneBench-Pro Public Release SubsetScience14.08.0−6.0
Concept Synth INDUCTION Challenge100 Round 1General QA31.025.0−6.0
ProofBenchMath77.071.0−6.0
WordleBenchGames97.091.0−6.0
MedXpertQA MM Extended (Mercor held-out)Health & Medicine56.951.0−5.9
ReactBench v1Code47.553.3+5.9
IntPhys 2Multimodal67.161.3−5.8
SciCode (main problems)Science21.515.8−5.7
CursorBench 3.2Code70.564.9−5.6
ChessBench AccuracyGames77.171.9−5.6
Artificial Analysis Agentic IndexIndexes52.847.4−5.4
RedlineBenchLegal42.447.8+5.4
ChessBench — Historical Output RepairGames46.040.7−5.3
ARC-AGI-2General QA89.283.9−5.3
BioMysteryBench - Human SolvableScience89.083.8−5.2
DTBenchGeneral QA98.493.3−5.1
TextQuests Average Cumulative HarmSafety23.722.2+5.0
CharXiv (descriptive and reasoning combined)Multimodal89.194.1+5.0
ZeroBench (Main questions, pass@5)Multimodal24.019.0−5.0
SAGEMath51.947.0−4.9
LMArena Agent Arena Bash ReliabilityAgents10.40.7−4.8
Concept Synth INDUCTION Challenge64 Round 1General QA42.237.5−4.7
Tax Agent BenchFinance & Accounting69.865.2−4.6
HealthBench Professional (Length-Adjusted)Health & Medicine60.956.3−4.6
LMArena Agent Arena Bash RecoveryAgents9.60.9−4.4
LMArena Agent Arena Steering BurdenAgents7.9-0.5−4.2
GPQA DiamondGeneral QA94.189.9−4.2
ERQAMultimodal67.863.6−4.2
AA-LCRSearch70.074.0+4.0
DAYJOB: HealthcareHealth & Medicine5.61.6−4.0
LiveCodeBench (Vals AI public dataset)Code89.885.9−3.8
Finance Agent v2 (Vals Index subset)Finance & Accounting56.352.6−3.7
Finance Agent v2 (Vals Multimodal Index component)Finance & Accounting56.352.6−3.7
LMArena Agent Arena SteerabilityAgents11.44.1−3.7
APEX-1 (Mercor, mean score)Professional Work66.069.5+3.5
LegalBenchLegal88.685.1−3.5
AA-LCR Extended (Mercor held-out)Search71.174.4+3.3
Ruby LLM Benchmarks — Program FixerCode77.274.0−3.3
GDPval (Mercor 213-task rubric score)Professional Work83.880.7−3.1
BioMysteryBench - Human DifficultScience46.549.4+2.9
Terminal-Bench Science 0.1 (Vals)Science12.910.0−2.9
APEX-Agents 1.1Professional Work74.571.8−2.7
HieroglyphBenchMultimodal22.920.5−2.4
AutomationBench 1.0.6 (Zapier)Agents17.114.8−2.3
GBENCH One-shot Coding GScoreGames65.663.3−2.2
AutomationBench (Overall)Agents17.415.2−2.2
BullshitBench v2 LeaderboardSafety56.054.0−2.0
ARC-AGI-1 Semi-PrivateGeneral QA98.596.5−2.0
ZeroBench Main Mean pass@1 (five samples)Multimodal15.814.2−1.6
MortgageTax (Vals all 1,258 documents)Finance & Accounting68.967.3−1.6
ObviousBenchGeneral QA99.397.9−1.4
FrontierMath Tiers 1-3 (v2) - PrivateMath87.086.0−1.1
Eval Connections One-Shot with TrapsGames85.086.0+1.0
ProgramBench (Mercor 100-task subset)Code0.01.0+1.0
ZeroBench Main Questions pass^5Multimodal9.010.0+1.0
ChartographyMultimodal34.834.0−0.8
TaxEval v2Finance & Accounting76.976.2−0.8
Machiavelli Deception (CAIS)Safety80.280.4−0.5
BrowseCompSearch88.087.5−0.5
CRED v0.92Science96.597.0+0.5
MASKSafety92.893.3+0.5
MedXpertQA MM (200-question set)Health & Medicine76.275.7−0.5
SUPERChemScience64.464.9+0.5
SWE-bench VerifiedCode95.095.4+0.4
CharXiv Extended (Mercor held-out)Multimodal80.781.1+0.4
SWE-bench MultilingualCode94.894.4−0.4
LMArena Agent Arena Tool HallucinationAgents0.41.0+0.3
OTIS Mock AIME 2024-2025Math100.099.7−0.3
Agents' Last Exam Overall ScoreAgents48.748.5−0.2
Artificial Analysis Coding IndexIndexes76.576.7+0.2
Terminal-Bench 2.1Code88.088.00.0
AIME 2025Math100.0100.00.0
Chess Puzzles (task version 2.0)Games47.047.00.0
Eval ConnectionsGames100.0100.00.0
SWE-bench Verified Extended (Mercor held-out)Code78.778.70.0

Claude Fable 5vsGPT-5.6 Terranormalised score marginsData Nov 15, 2026Data: Mercor, Vals AI, Center for AI Safety and 51 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…