Atlas

Compare Models

Claude Fable 5 vs Muse Spark 1.3

Claude Fable 5 leads by 2.3 capability points (≈95% interval +0.0 to +4.6): a clear lead over Muse Spark 1.3 · P(higher) ≈ 98%

Wins 81 of 111 shared benchmarks (28 lost, 2 tied)

Capability vs. relative cost
Claude Fable 5Muse Spark 1.3Closed 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 Spark 1.3capability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins81 – 28 · 111 shared, 2 tied

Comparing matching setups only (69 benchmarks excluded) · suites counted once
← Claude Fable 5Muse Spark 1.3 →

Claude Fable 5vsMuse Spark 1.3wins by categoryData Nov 15, 2026Data: Mercor, Arena, Surge AI and 35 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Muse Spark 1.3 minus Claude Fable 5.

KernelBench Mega Normalized (H100)Code78.712.4−66.3
Olam Social Poker EloGames15531494−64.1
BabaIsBench Stage 0: The IntroGames100.046.0−54.0
FrontierMath Tier 4 (v2) - PrivateMath90.246.3−43.9
ProofBench v1.1Math95.058.0−37.0
VoxelBench Text RatingGames21011616−34.5
RiemannBenchMath60.028.0−32.0
Mystery Game PuzzlesGames52.025.0−27.0
LLM Creative Story WritingChat & Writing2.90.6−25.9
SWE-bench Verified Extended (Mercor held-out)Code78.754.0−24.7
TaxCalcBench TY25Finance & Accounting30.06.0−24.0
WordleBenchGames97.073.0−24.0
TextQuests Game ProgressGames56.133.0−23.1
KernelBench CUDA Normalized (RTX PRO 6000)Code51.630.3−21.3
Document ArenaProfessional Work14971473−21.2
Medical Long Context Reasoning (MLCR-AA)Health & Medicine64.443.3−21.1
HANDBOOK.mdProfessional Work36.215.4−20.8
APEX-SWECode58.939.1−19.8
LLM Debate BenchmarkChat & Writing17271593−19.0
ALE-BenchCode24631981−17.1
Terminal-Bench 4.0Code44.527.8−16.8
Humanity's Last ExamGeneral QA57.843.2−14.6
ArXivMath 05/2026Math87.573.3−14.2
ComplexConstraintsChat & Writing38.151.9+13.8
LMArena Agent Arena Net ImprovementAgents8.03.8−13.7
APEX-Agents 1.1Professional Work74.561.3−13.2
GBENCH One-shot Coding GScoreGames65.652.7−12.8
FrontierMath Tiers 1-3 (v2) - PrivateMath87.074.4−12.6
GeneBench-Pro Public Release SubsetScience14.026.0+12.0
Humanity's Last Exam DiamondGeneral QA45.633.7−11.9
NoCoT-Bench NCRI 15.2General QA11998.3−11.4
AA-AnalystAgent (pass^5)Professional Work48.837.5−11.2
NoCoT-Bench NCKI (kspine v3)General QA128113−10.7
HiL-Bench (Human-in-Loop Benchmark)Agents56.366.9+10.6
Chess Puzzles (task version 2.0)Games47.037.0−10.0
ACCoRDGeneral QA85.075.1−9.9
LMArena Agent Arena Net Improvement (Steering Burden)Agents7.32.2−9.7
GDPval Extended (Mercor held-out rubric score)Professional Work65.355.9−9.4
ChessBench EloGames10461201+9.2
Terminal-Bench 4.0 (Artificial Analysis)Code42.433.3−9.1
Conceptual Reasoning IndexIndexes72.766.2−8.9
LiveBench 2.0Agents83.674.8−8.8
ArXivMath 06/2026Math85.476.7−8.7
ChessBench CoherenceGames82.691.2+8.6
EnterpriseBench: CoreCraftProfessional Work70.361.9−8.4
BrokenArXiv 05/2026Math44.536.5−8.0
Extended NYT Connections 940 (Quadratic v1)Games92.785.1−7.6
MineBench — Bradley-Terry RatingGames18691747−7.4
LMCAGeneral QA61.153.9−7.3
ChartographyMultimodal34.827.6−7.2
APEX-1 (Mercor, mean score)Professional Work66.073.0+7.0
Design Arena Overall Code CategoriesCode13061348+7.0
HardGeoBench Avg DistanceMultimodal8001022−6.8
GameBench 2Games75.268.5−6.7
SWE-TogetherCode68.862.4−6.4
LMArena Agent Arena Praise vs ComplaintAgents14.31.7−6.3
Excel Modeling BenchmarkFinance & Accounting73.767.4−6.2
OSWorld 2.0 - reported partial reward, patch and step limit unspecifiedAgents72.966.9−6.0
ArXivMath 04/2026Math72.566.7−5.8
DAYJOB: FinanceFinance & Accounting9.014.8+5.8
Vals RSI Index v1.1 (historical five tasks)Agents24.518.7−5.8
DeepSWE v1.1Code69.975.4+5.5
BrokenArXiv 06/2026Math47.853.2+5.4
LMArena Agent Arena SteerabilityAgents11.41.6−4.9
BrokenArXiv 04/2026Math54.149.2−4.9
CharXiv (descriptive and reasoning combined)Multimodal89.193.9+4.8
Vals RSI Index v1.1 (four tasks)Agents24.319.6−4.7
Vision Arena OverallMultimodal13091291−4.6
Vibe Code Bench v1.1Code90.485.9−4.5
LMArena Agent Arena Bash ReliabilityAgents10.42.0−4.2
AA-LCR Extended (Mercor held-out)Search71.166.9−4.2
DeepSearchQA (F1 score)Search94.490.3−4.1
OSWorld 2.0 - reported binary completion, patch and step limit unspecifiedAgents36.132.0−4.1
CadQueryEval (Corrected Geometry)Code80.084.0+4.0
LMArena Code Arena WebDevCode16251655+3.9
MMMU-Pro Extended (Mercor held-out)Multimodal37.633.8−3.8
Arena Image-to-WebDevCode16211637+3.7
AutomationBench 1.0.6 (Zapier)Agents17.120.7+3.6
AA-Briefcase EloAgents15301589+3.5
MedXpertQA MM Extended (Mercor held-out)Health & Medicine56.960.1+3.2
Ruby LLM Benchmarks — Program FixerCode77.274.1−3.2
Bug Hunt BenchCode27.630.7+3.0
GDPval-AA v2Professional Work17231662−3.0
Humanity's Sixth SenseMultimodal34.537.4+2.9
APEX-AccountingFinance & Accounting56.453.5−2.9
LMArena Agent Arena Steering BurdenAgents7.92.6−2.6
Tax Agent BenchFinance & Accounting69.872.4+2.6
HardGeoBench <=10miMultimodal41.338.8−2.5
DAYJOB: HealthcareHealth & Medicine5.67.8+2.2
GDP.pdfProfessional Work29.827.6−2.2
Harvey LAB-AA v1.0 (criterion pass rate)Legal93.695.5+1.9
DTBenchGeneral QA98.496.5−1.9
LMArena Agent Arena Bash RecoveryAgents9.66.1−1.8
Artificial Analysis Intelligence Index v4.3Indexes49.748.2−1.5
Terminal-Bench Science 0.1 (Vals)Science12.914.3+1.4
LMArena Text Arena EloChat & Writing15041496−1.3
GPQA DiamondGeneral QA92.691.3−1.3
HardGeoBench Country AccuracyMultimodal61.360.0−1.3
HardGeoBench <=100miMultimodal56.355.0−1.2
ChessBench AccuracyGames77.176.0−1.2
CharXiv Extended (Mercor held-out)Multimodal80.781.8+1.1
Roboflow Vision Evals (September 2026 protocol)Multimodal78.779.8+1.1
Eval Connections One-Shot with TrapsGames85.084.0−1.0
PCT Political Inconsistency (CAIS)Safety39.840.6−0.9
OTIS Mock AIME 2024-2025Math100.099.2−0.8
Terminal-Bench 2.1Code88.088.8+0.8
LMArena Agent Arena Confirmed SuccessAgents5.13.7−0.7
LMArena Agent Arena Tool HallucinationAgents0.40.8+0.2
MedXpertQA MM (200-question set)Health & Medicine76.276.3+0.1
AIME 2025Math100.0100.00.0
ObviousBenchGeneral QA99.399.30.0

Claude Fable 5vsMuse Spark 1.3normalised score marginsData Nov 15, 2026Data: Mercor, Arena, Surge AI and 35 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…