Atlas

Compare Models

Claude Fable 5.1 vs Claude Fable 5.2

Claude Fable 5.2 leads by 4.5 capability points (≈95% interval +1.7 to +7.3): a clear lead over Claude Fable 5.1 · P(higher) ≈ 100%

Wins 66 of 83 shared benchmarks (16 lost, 1 tied)

Capability vs. relative cost
Claude Fable 5.1Claude Fable 5.2Closed 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 5.1vsClaude Fable 5.2capability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins16 – 66 · 83 shared, 1 tied

Comparing matching setups only (6 benchmarks excluded) · suites counted once
← Claude Fable 5.1Claude Fable 5.2 →

Claude Fable 5.1vsClaude Fable 5.2wins by categoryData Nov 15, 2026Data: Artificial Analysis, Epoch AI, Arena and 25 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Claude Fable 5.2 minus Claude Fable 5.1.

AA-PersonalityBench Informativeness indexChat & Writing96.246.7−50.0
HoneyBench v0.2 (hack rate)Safety61.811.9+49.9
Anthropic ECI (AECI), November 2026 fitIndexes173177+39.4
BioSecBench-Structure V1 (58 evaluations)Science5.843.7+37.9
EBR-bench v4 (card ban, single agent)Games57.190.5+33.3
Mystery Game PuzzlesGames58.090.0+32.0
Chess Puzzles (task version 2.0)Games48.079.0+31.0
LMArena Agent Arena Net Improvement (Steering Burden)Agents14.830.4+29.3
Clocktower BenchGames15671759+27.5
ChartographyMultimodal46.273.4+27.2
Humanity's Last ExamGeneral QA46.573.2+26.7
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine18.343.9+25.6
OpenWorldEngineer (Mean@8)Code29.154.6+25.4
AA-PersonalityBench Conviction indexChat & Writing10075.3−25.0
ChessBenchGames45.369.3+24.0
Remote Labor Index (RLI)Professional Work17.941.3+23.4
HieroglyphBenchMultimodal20.443.6+23.2
SimpleQA Verified (Epoch anti-abstention)General QA70.894.0+23.2
AutomationBench 1.0.6 (Zapier)Agents22.444.3+21.9
LMArena Agent Arena Praise vs ComplaintAgents31.673.4+20.9
Furniture Assembly BenchmarkMultimodal70.090.0+20.0
AA-PersonalityBench Imagination indexChat & Writing86.767.6−19.2
AA-3D v1.0 EloCode11391271+19.2
SpireClimbGames2.019.6+17.6
AA-PersonalityBench Slop indexChat & Writing23.86.7−17.3
AA-PersonalityBench Expressiveness indexChat & Writing98.181.0−17.3
AA-PersonalityBench Transparency indexChat & Writing98.181.0−17.3
FrontierCode 2.0 - Pass RateCode37.654.7+17.1
FrontierCode 2.0 - ScoreCode34.050.2+16.2
Bug Hunt BenchCode41.055.2+14.3
LMArena Code Arena WebDevCode17421836+12.3
FrontierMath Tier 4 (v2) - PrivateMath87.8100.0+12.2
AA-PersonalityBench Warmth indexChat & Writing94.382.9−11.5
InferenceBenchCode9.310.3+11.5
HardGeoBench Country AccuracyMultimodal58.870.0+11.3
Tax Agent BenchFinance & Accounting77.967.0−10.9
APEX-Agents 1.1Professional Work77.187.6+10.5
LMArena Agent Arena Steering BurdenAgents14.833.0+9.1
HardGeoBench Avg DistanceMultimodal845552+9.1
GDPxProfessional Work8.516.4+7.9
Harvey LAB (Mercor run, undisclosed ~213-task cohort)Legal17.410.0−7.4
FrontierCode 1.1 Main (Artificial Analysis)Code49.056.2+7.2
DeepSWE v1.1Code67.374.3+7.0
Conceptual Reasoning IndexIndexes74.979.9+6.8
Agent Initiative BenchmarkAgents7.58.2+6.7
AA-Briefcase v1.1 EloAgents16711785+6.3
HardGeoBench <=100miMultimodal53.860.0+6.3
CritPt v2 (agentic, avg@5)Science77.783.7+6.0
Epoch Capabilities Index (ECI)Indexes167174+5.8
AA-PersonalityBench Organization indexChat & Writing60.054.3−5.8
ACCoRDGeneral QA83.789.4+5.7
SWE-bench MultilingualCode91.196.6+5.5
LMCAGeneral QA65.570.6+5.2
GameBench 2Games94.099.1+5.1
FrontierMath Tiers 1-3 (v2) - PrivateMath90.295.1+4.9
SciCode 1.0.1 (Artificial Analysis)Code63.167.7+4.6
CritPt (avg@5)Science31.135.7+4.6
AA-LCR v1.1Search87.583.3−4.2
MysteryMechanismScience47.751.8+4.1
DualEntry Accounting AI BenchmarkFinance & Accounting85.181.2−3.9
LMArena Agent Arena Confirmed SuccessAgents14.422.0+3.8
HealthBenchHealth & Medicine63.960.1−3.8
ARC-AGI-2General QA90.093.8+3.8
HardGeoBench <=10miMultimodal45.048.8+3.8
APEX-Accounting 1.1Finance & Accounting66.770.4+3.7
DAXBench v2Code91.588.3−3.2
HealthBench (length-adjusted, Artificial Analysis)Health & Medicine57.260.4+3.2
GDPval-AA v2.1Professional Work17651831+2.9
Ramp Accounting BenchFinance & Accounting49.752.2+2.5
HealthBench ProfessionalHealth & Medicine69.271.4+2.2
LMArena Agent Arena Bash ReliabilityAgents10.114.4+2.2
Vals RSI Index v1.1 (four tasks)Agents36.137.9+1.8
τ³-Banking (Mercor run, bm25_grep retrieval)Agents54.055.7+1.7
DTBenchGeneral QA97.699.2+1.6
Immigration BenchLegal75.474.0−1.5
BuseyBench SVGMultimodal7.07.2+1.4
ARC-AGI-1 Semi-PrivateGeneral QA97.598.5+1.0
Harvey LAB-AA v1.1 (hallucination-gated)Legal6.47.2+0.8
Excel Modeling Benchmark v1.1Finance & Accounting76.777.4+0.6
OTIS Mock AIME 2024-2025Math100.099.7−0.3
MineBench — Bradley-Terry RatingGames19281932+0.2
LMArena Agent Arena Tool HallucinationAgents0.81.0+0.1
ProofBench v1.1Math100.0100.00.0

Claude Fable 5.1vsClaude Fable 5.2normalised score marginsData Nov 15, 2026Data: Artificial Analysis, Epoch AI, 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…