Atlas

Compare Models

Claude Fable 5 vs Muse Wonder

Muse Wonder leads by 0.1 capability points (≈95% interval −2.3 to +2.5): statistically tied · P(higher) ≈ 53%

Claude Fable 5 wins 40 of 78 shared benchmarks (37 lost, 1 tied)

Capability vs. relative cost
Claude Fable 5Muse WonderClosed 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 5vsMuse Wondercapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins40 – 37 · 78 shared, 1 tied

Comparing matching setups only (53 benchmarks excluded) · suites counted once
← Claude Fable 5Muse Wonder →

Claude Fable 5vsMuse Wonderwins by categoryData Nov 15, 2026Data: Center for AI Safety, Mercor, Arena and 25 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Muse Wonder minus Claude Fable 5.

LisanBench Path LengthGames1368027973+44.4
LisanBench Difficulty-WeightedGames45628303+40.1
VCT-Refusal (CAIS)Safety100.060.2−39.8
Vending-Bench 2Games45309432+29.7
HiL-Bench (Human-in-Loop Benchmark)Agents56.379.7+23.3
FrontierMath Tier 4 (v2) - PrivateMath90.270.7−19.5
Furniture Assembly BenchmarkMultimodal35.854.2+18.3
ChessBench CoherenceGames82.6100.0+17.4
VoxelBench Text RatingGames21011870−16.5
Design Arena Overall Code CategoriesCode13061385+13.1
NoCoT-Bench NCKI (kspine v3)General QA128109−13.0
DAYJOB: HealthcareHealth & Medicine5.618.0+12.4
DAYJOB: FinanceFinance & Accounting9.021.3+12.3
MMMU ProMultimodal72.784.9+12.2
Eval Connections One-Shot with TrapsGames85.073.0−12.0
NoCoT-Bench NCRI 15.2General QA11997.7−11.7
ProofBench v1.1Math95.084.0−11.0
LMArena Agent Arena Net Improvement (Steering Burden)Agents7.31.8−10.5
Humanity's Last ExamGeneral QA57.847.6−10.2
MMLU ProGeneral QA79.489.6+10.2
Machiavelli Deception (CAIS)Safety80.283.6−10.2
LLM Creative Story WritingChat & Writing2.92.0−10.1
TextQuests Average Cumulative HarmSafety23.720.8+10.1
SpatialViz-BenchMultimodal73.983.5+9.6
Humanity's Sixth SenseMultimodal34.543.7+9.2
APEX-1 (Mercor, mean score)Professional Work66.075.1+9.1
DeepSWE v1.1Code69.978.8+8.9
EnterpriseBench: CoreCraftProfessional Work70.361.5−8.8
ACCoRDGeneral QA85.076.2−8.8
TextQuests Game ProgressGames56.147.6−8.4
LMArena Agent Arena Praise vs ComplaintAgents14.3-1.8−8.0
TaxCalcBench TY25Finance & Accounting30.022.0−8.0
MindCube TinyMultimodal83.191.1+8.0
MASKSafety92.884.9−7.9
ERQAMultimodal67.875.3+7.5
GBENCH One-shot Coding GScoreGames65.658.1−7.5
ChessBench AccuracyGames77.170.2−7.4
Vibe Code Bench v1.1Code90.483.1−7.3
Conceptual Reasoning IndexIndexes72.767.7−6.8
AA-Briefcase v1.1 EloAgents15361659+6.7
Humanity's Last Exam DiamondGeneral QA45.639.0−6.6
Excel Modeling BenchmarkFinance & Accounting73.767.1−6.6
PCT Political Inconsistency (CAIS)Safety39.833.3+6.5
APEX-Accounting 1.1Finance & Accounting61.067.1+6.1
SciCode (main problems)Science21.515.8−5.7
MMMU-Pro Extended (Mercor held-out)Multimodal37.643.2+5.6
LiveBench 2.0Agents83.678.1−5.5
HealthBench Professional (Length-Adjusted)Health & Medicine66.060.5−5.5
CharXiv (descriptive and reasoning combined)Multimodal89.194.5+5.4
FrontierMath Tiers 1-3 (v2) - PrivateMath87.082.1−4.9
GDPval Extended (Mercor held-out rubric score)Professional Work65.370.0+4.7
LMCAGeneral QA61.156.8−4.3
APEX-SWECode58.954.6−4.3
Terminal-Bench 4.0 (Artificial Analysis)Code42.438.4−4.0
AA-LCR Extended (Mercor held-out)Search71.174.4+3.3
DTBenchGeneral QA98.495.2−3.2
GDPval (Mercor 213-task rubric score)Professional Work83.887.0+3.2
IntPhys 2Multimodal67.170.2+3.1
LMArena Code Arena WebDevCode16251649+3.0
HPCT Refusal (CAIS)Safety100.097.0−3.0
LMArena Agent Arena Bash ReliabilityAgents10.44.8−2.8
LMArena Agent Arena Steering BurdenAgents7.92.6−2.7
GameBench 2Games75.277.8+2.6
Bug Hunt BenchCode27.630.2+2.5
APEX-Agents 1.1Professional Work74.577.0+2.5
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine20.622.2+1.7
Tax Agent BenchFinance & Accounting69.871.2+1.3
BioTIER RefusalSafety100.098.7−1.3
LMArena Text Arena EloChat & Writing15041511+1.2
CharXiv Extended (Mercor held-out)Multimodal80.779.6−1.1
Mystery Game PuzzlesGames52.051.0−1.0
BullshitBench v2 LeaderboardSafety56.057.0+1.0
OTIS Mock AIME 2024-2025Math100.099.4−0.6
LMArena Agent Arena Tool HallucinationAgents0.4-0.5−0.5
Prophet Arena Arena Score (Fixed Context)Forecasting84.985.2+0.3
LMArena Agent Arena Confirmed SuccessAgents5.15.4+0.1
MineBench — Bradley-Terry RatingGames18691868−0.1
ChessBench EloGames104610460.0

Claude Fable 5vsMuse Wondernormalised score marginsData Nov 15, 2026Data: Center for AI Safety, Mercor, Arena and 25 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…