Atlas

Compare Models

Claude Fable 5.1 vs Opus 4.8

Claude Fable 5.1 leads by 6.0 capability points (≈95% interval +3.8 to +8.2): a clear lead over Opus 4.8 · P(higher) ≈ 100%

Wins 164 of 180 shared benchmarks (14 lost, 2 tied)

Capability vs. relative cost
Claude Fable 5.1Opus 4.8Closed 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 5.1vsOpus 4.8capability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins164 – 14 · 180 shared, 2 tied

Comparing matching setups only (23 benchmarks excluded) · suites counted once
← Claude Fable 5.1Opus 4.8 →

Claude Fable 5.1vsOpus 4.8wins by categoryData Nov 15, 2026Data: Vals AI, Mercor, Center for AI Safety and 60 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Opus 4.8 minus Claude Fable 5.1.

VCT-Refusal (CAIS)Safety98.941.2−57.7
BrokenArXiv 05/2026Math78.023.0−55.0
Clocktower BenchGames15671195−53.4
Time Horizon Index: KSPAgents63.311.8−51.5
BrokenArXiv 06/2026Math86.641.7−44.9
GSO Opt@1Code88.247.1−41.2
TasteVal (compute multiplier)Science1.70.8−38.4
Document ArenaProfessional Work15151475−36.0
Terminal-Bench 4.0Code57.923.6−34.2
LLM Creative Story WritingChat & Writing3.80.8−33.9
FrontierMath Tier 4 (v2) - PrivateMath87.856.1−31.7
GameBench 2Games94.062.5−31.6
Concept Synth INDUCTION Challenge64 Round 1General QA34.46.3−28.1
Furniture Assembly BenchmarkMultimodal70.042.5−27.5
Blueprint-Bench 2Multimodal0.40.1−27.4
KernelBench CUDA Normalized (RTX PRO 6000)Code51.624.9−26.7
Bug Hunt BenchCode41.014.3−26.7
ALE-BenchCode28492102−26.5
HiL-Bench (Human-in-Loop Benchmark)Agents61.535.3−26.2
Olam Social Poker EloGames14941518+26.1
Medical Long Context Reasoning (MLCR-AA)Health & Medicine71.145.6−25.6
EnterpriseBench: CoreCraftProfessional Work77.452.3−25.1
LMArena Code Arena WebDevCode17421555−24.4
Terminal-Bench Science 0.1 (Vals)Science34.310.0−24.3
KernelBench Mega Normalized (RTX PRO 6000)Code64.740.6−24.1
LMArena Agent Arena Net ImprovementAgents14.16.9−23.8
EnigmaEvalGeneral QA44.220.5−23.7
OSWorld 2.0 - reported partial reward, patch and step limit unspecifiedAgents77.954.8−23.1
ArXivMath 05/2026Math87.565.0−22.5
TaxCalcBench TY25Finance & Accounting42.020.0−22.0
Mystery Game PuzzlesGames58.036.0−22.0
SimpleBenchGeneral QA86.664.8−21.8
OSWorld 2.0 - reported binary completion, patch and step limit unspecifiedAgents41.720.6−21.1
ArXivMath 06/2026Math91.070.0−21.0
FrontierCode 2.0 - Pass RateCode37.616.7−20.9
TextQuests Average Cumulative HarmSafety27.121.0+20.7
PCT Political Inconsistency (CAIS)Safety33.453.8−20.4
LisanBench Difficulty-WeightedGames45732694−20.1
AA-Briefcase EloAgents16621316−20.1
APEX-SWECode63.643.9−19.7
MindCube TinyMultimodal84.564.9−19.6
FrontierCode 2.0 - ScoreCode34.014.9−19.2
RiemannBenchMath65.647.2−18.4
BuseyBench SVGMultimodal7.05.2−18.2
BabyVisionMultimodal46.428.4−18.0
ARC-AGI-2General QA90.072.1−17.9
ChessBench CoherenceGames61.979.7+17.8
SimpleQA Verified (Epoch anti-abstention)General QA70.853.0−17.8
LisanBench Path LengthGames151009463−17.5
HardGeoBench <=100miMultimodal53.836.3−17.5
HardGeoBench <=10miMultimodal45.027.5−17.5
SWE Atlas - Test WritingCode67.049.6−17.4
BullshitBench v2 LeaderboardSafety77.094.0+17.0
ReactBench (39 tasks)Code49.232.3−16.9
HANDBOOK.mdProfessional Work38.821.9−16.9
TextQuests Game ProgressGames56.640.1−16.5
Roboflow Vision Evals (October 2026, five tasks)Multimodal74.758.3−16.4
RuneBenchGames6.15.1−16.3
MMMU-Pro Extended (Mercor held-out)Multimodal47.631.4−16.2
SM BenchSafety89.072.9−16.1
Hemingway-benchChat & Writing10951038−15.8
GBENCH One-shot Coding GScoreGames69.354.0−15.3
LLM Debate BenchmarkChat & Writing17381631−15.2
ArXivMath 04/2026Math77.562.5−15.0
HPCT Refusal (CAIS)Safety100.085.0−15.0
SUPERChemScience74.760.3−14.4
MedXpertQA MM Extended (Mercor held-out)Health & Medicine66.051.6−14.4
NoCoT-Bench NCRI 15.2General QA128102−14.3
SWE Atlas Codebase QnA Extended (Mercor held-out)Code38.024.0−14.0
GDPval-AA v2.1Professional Work17651461−13.4
InteractiveBench (task pass)Agents41.728.3−13.3
Terminal-Bench 2.1 Extended (Mercor held-out)Code19.532.7+13.2
Design Arena Overall Code CategoriesCode13311252−13.1
APEX-AccountingFinance & Accounting61.048.0−13.0
Tax Agent BenchFinance & Accounting77.964.9−13.0
DeepSWE v1.1Code67.354.4−12.9
ActiveVisionMultimodal15.32.4−12.9
Agent Red Teaming k=100 (CAIS)Safety34.247.0−12.8
SpatialViz-BenchMultimodal77.965.2−12.7
APEX-Agents 1.1Professional Work77.164.4−12.7
IntPhys 2Multimodal67.755.1−12.6
Roboflow Vision Evals (September 2026 protocol)Multimodal81.368.7−12.6
Antidote: Everyday EditionSafety10651030−12.5
AA-AnalystAgent (pass^5)Professional Work57.545.0−12.5
LVBenchMultimodal79.767.3−12.4
HardGeoBench Avg DistanceMultimodal8451245−12.3
Vals RSI Index v1.1 (four tasks)Agents36.124.0−12.1
Vals RSI Index v1.1 (historical five tasks)Agents35.023.2−11.8
GDPval-AA v2Professional Work17641529−11.5
EurekaBenchScience29.818.4−11.4
Artificial Analysis Intelligence Index v4.3Indexes53.442.0−11.4
LiveBench 2.0Agents88.076.8−11.2
MedXpertQA MM (200-question set)Health & Medicine83.772.5−11.2
APEX-1 (Mercor, mean score)Professional Work73.862.7−11.1
FrontierFinance v3.1Finance & Accounting55.945.0−10.8
ERQAMultimodal70.359.5−10.8
Finance Agent v2 - General Qualitative AnalysisFinance & Accounting77.066.4−10.5
AdvancedIFChat & Writing84.273.7−10.5
DeepResearch Bench II (Mercor 130-task set)Search55.044.6−10.4
FrontierMath Tiers 1-3 (v2) - PrivateMath90.280.0−10.2
WeirdML (v2)Science92.982.9−10.0
GDPval Extended (Mercor held-out rubric score)Professional Work68.758.9−9.8
AA-LCR Extended (Mercor held-out)Search73.063.2−9.8
Remote Labor Index (RLI)Professional Work17.98.3−9.6
ComplexConstraintsChat & Writing45.135.6−9.5
Conceptual Reasoning IndexIndexes74.968.3−9.2
HealthBench ProfessionalHealth & Medicine69.260.3−8.9
Drone-BenchAgents91.282.3−8.9
SciCode 1.0.1 (Artificial Analysis)Code63.154.4−8.7
MineBench — Bradley-Terry RatingGames19281786−8.6
MASKSafety85.994.4+8.5
Finance Agent v2 - General Quantitative AnalysisFinance & Accounting79.771.6−8.1
CadQueryEval (Corrected Geometry)Code76.068.0−8.0
GeneBench-Pro Public Release SubsetScience26.018.0−8.0
ChessBench EloGames1095961−8.0
LMCAGeneral QA65.557.5−8.0
SWE-TogetherCode69.361.5−7.8
NoCoT-Bench NCKI (kspine v3)General QA128117−7.7
SciCode (main problems)Science24.216.9−7.3
Excel Modeling BenchmarkFinance & Accounting76.769.4−7.3
Gray Swan IPI Attack Success (k=15)Safety1.08.0−7.0
Finance Agent v2 - PrecedentsFinance & Accounting32.225.6−6.6
Toloka Arena (v1)Agents60.754.3−6.4
HieroglyphBenchMultimodal20.414.0−6.4
ACCoRDGeneral QA83.777.4−6.3
SAGEMath48.554.8+6.3
Finance Agent v2 - AdjustmentsFinance & Accounting54.448.4−6.0
Epoch Capabilities Index (ECI)Indexes167160−5.8
Arena Alignment IndexSafety80.174.5−5.6
MedScribeHealth & Medicine91.385.8−5.5
CharXiv Extended (Mercor held-out)Multimodal82.276.7−5.5
AutomationBench 1.0.6 (Zapier)Agents22.416.9−5.5
HealthBenchHealth & Medicine63.958.8−5.1
ARC-AGI-1 Semi-PrivateGeneral QA97.592.5−5.0
HardGeoBench Country AccuracyMultimodal58.853.8−5.0
LegalBenchLegal88.583.6−4.9
BioTIER RefusalSafety99.594.7−4.8
BrowseComp (Mercor 130-question subset)Search78.373.7−4.6
Finance Agent v2 - Disclosure AnalysisFinance & Accounting70.365.7−4.6
Finance Agent v2 - All PassFinance & Accounting47.843.5−4.3
Senior SWE-BenchCode34.730.5−4.2
MMMU-Pro (Vals 4-Option)Multimodal90.686.6−4.0
Machiavelli Deception (CAIS)Safety82.483.7−4.0
BabaIsBench Stage 0: The IntroGames100.096.0−4.0
Chess Puzzles (task version 2.0)Games48.052.0+4.0
LMArena Text Arena EloChat & Writing15031482−3.7
GeoBenchMultimodal43034128−3.5
Finance Agent v2 - Earnings AnalysisFinance & Accounting75.572.1−3.4
Ruby LLM Benchmarks — Program FixerCode76.072.6−3.3
Humanity's Last ExamGeneral QA46.549.8+3.3
MMLU ProGeneral QA93.090.0−3.0
SpeciEval (Pooled English Runs)Safety87.284.3−2.9
SpeciEval (Refusal-Aware Retries, Historical)Safety87.284.3−2.9
SWE Atlas - Codebase QnACode60.057.3−2.7
DTBenchGeneral QA97.694.9−2.7
FrontierCode 1.1 Extended - ScoreCode62.059.6−2.4
FrontierCode 1.1 Extended - Pass RateCode67.965.5−2.4
Finance Agent v2 - Market AnalysisFinance & Accounting69.567.2−2.3
LMArena Agent Arena Bash RecoveryAgents11.67.2−2.2
GDPval (Mercor 213-task rubric score)Professional Work86.684.5−2.1
LiveCodeBench (Vals Public) - MediumCode91.989.8−2.1
SWE-bench Verified Extended (Mercor held-out)Code61.363.3+2.0
Finance Agent v2 - Financial ModelingFinance & Accounting25.323.3−2.0
CharXiv (descriptive and reasoning combined)Multimodal95.193.2−1.9
AIME 2025Math100.098.3−1.7
OTIS Mock AIME 2024-2025Math100.098.3−1.7
GlobeBenchGeneral QA21.620.0−1.6
ChessBench AccuracyGames71.372.4+1.2
Finance Agent v2 - ComparablesFinance & Accounting46.044.9−1.1
Extended NYT Connections 940 (Quadratic v1)Games90.091.1+1.1
GPQA DiamondGeneral QA92.991.9−1.0
MortgageTax (Vals all 1,258 documents)Finance & Accounting70.869.9−0.9
Vision Arena OverallMultimodal12891286−0.9
LMArena Agent Arena SteerabilityAgents8.49.6+0.6
Terminal-Bench 2.1Code81.682.0+0.4
TaxEval v2Finance & Accounting76.075.6−0.3
MedCodeHealth & Medicine53.553.2−0.3
DistressBenchSafety71.571.4−0.1
LiveCodeBench (Vals Public) - EasyCode99.199.10.0
ObviousBenchGeneral QA99.399.30.0

Claude Fable 5.1vsOpus 4.8normalised score marginsData Nov 15, 2026Data: Vals AI, Mercor, Center for AI Safety and 60 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…