Atlas

Compare Models

Claude Fable 5.2 vs Claude Sonnet 5.5

Claude Fable 5.2 leads by 3.6 capability points (≈95% interval +0.8 to +6.5): a clear lead over Claude Sonnet 5.5 · P(higher) ≈ 99%

Wins 66 of 80 shared benchmarks (13 lost, 1 tied)

Capability vs. relative cost
Claude Fable 5.2Claude Sonnet 5.5Closed 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 5.2vsClaude Sonnet 5.5capability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins66 – 13 · 80 shared, 1 tied

Comparing matching setups only (2 benchmarks excluded) · suites counted once
← Claude Fable 5.2Claude Sonnet 5.5 →

Claude Fable 5.2vsClaude Sonnet 5.5wins by categoryData Nov 15, 2026Data: Artificial Analysis, Epoch AI, Arena and 22 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Claude Sonnet 5.5 minus Claude Fable 5.2.

SimpleQA Verified (Epoch anti-abstention)General QA94.046.5−47.5
AA-PersonalityBench Informativeness indexChat & Writing46.784.8+38.5
AA-PersonalityBench Organization indexChat & Writing54.384.8+30.8
ChessBenchGames69.339.3−30.0
HoneyBench v0.2 (hack rate)Safety11.941.6−29.7
LMArena Agent Arena Net Improvement (Steering Burden)Agents30.416.0−27.2
Humanity's Last ExamGeneral QA73.246.1−27.1
Humanity's Last Exam DiamondGeneral QA80.253.3−26.9
Mystery Game PuzzlesGames90.065.0−25.0
ARC-AGI-3 (Semi-Private)General QA97.372.4−24.8
BioSecBench-Structure V1 (58 evaluations)Science43.721.8−21.8
LMArena Agent Arena Praise vs ComplaintAgents73.430.5−21.5
InferenceBenchCode10.312.1+21.4
AA-PersonalityBench Slop indexChat & Writing6.727.7+21.2
Chess Puzzles (task version 2.0)Games79.059.0−20.0
FrontierMath Tier 4 (v2) - PrivateMath100.080.5−19.5
AA-PersonalityBench Transparency indexChat & Writing81.0100+19.2
SpireClimbGames19.61.3−18.3
AA-PersonalityBench Imagination indexChat & Writing67.684.8+17.3
AA-PersonalityBench Conviction indexChat & Writing75.358.1−17.3
OpenWorldEngineer (Mean@8)Code54.639.0−15.6
Agent Initiative BenchmarkAgents8.26.7−15.2
Furniture Assembly BenchmarkMultimodal90.075.0−15.0
HardGeoBench <=10miMultimodal48.836.3−12.5
ChartographyMultimodal73.461.6−11.8
HardGeoBench Country AccuracyMultimodal70.058.8−11.2
CritPt v2 (agentic, avg@5)Science86.375.7−10.6
FrontierCode 1.1 Main (Artificial Analysis)Code56.245.8−10.4
HardGeoBench <=100miMultimodal60.050.0−10.0
ARC-AGI-2General QA93.883.9−9.9
AA-3D v1.0 EloCode12711204−9.8
Ramp Accounting BenchFinance & Accounting52.242.5−9.7
AA-PersonalityBench Warmth indexChat & Writing82.973.3−9.6
LMArena Agent Arena Steering BurdenAgents33.014.5−9.2
LMArena Code Arena WebDevCode18361773−8.3
GDPxProfessional Work16.48.2−8.2
FrontierSWE V2Code69.961.9−8.0
AA-PersonalityBench Expressiveness indexChat & Writing81.073.3−7.7
GameBench 2Games99.191.4−7.7
SciCode 1.0.1 (Artificial Analysis)Code67.761.0−6.7
Epoch Capabilities Index (ECI)Indexes174166−6.4
BuseyBench SVGMultimodal7.26.5−6.4
FrontierCode 2.0 - Pass RateCode52.746.4−6.4
Bug Hunt BenchCode55.248.9−6.3
FrontierMath Tiers 1-3 (v2) - PrivateMath95.188.8−6.3
FrontierCode 2.0 - ScoreCode48.242.2−6.0
APEX-Accounting 1.1Finance & Accounting70.464.5−5.9
HealthBenchHealth & Medicine60.165.9+5.8
SWE-bench MultilingualCode96.690.9−5.7
HLAB-Vals-1.1Legal5.010.4+5.4
Immigration BenchLegal74.069.0−5.0
CritPt (avg@5)Science35.731.4−4.3
HardGeoBench Avg DistanceMultimodal552683−4.1
MineBench — Bradley-Terry RatingGames19321998+4.0
SWE-bench Multimodal (Vals v1, 480 Tasks)Code62.358.3−4.0
APEX-Agents 1.1Professional Work87.683.7−3.9
Conceptual Reasoning IndexIndexes79.977.1−3.9
DeepSWE v1.1Code74.370.5−3.8
ACCoRDGeneral QA89.485.7−3.7
AA-LCR v1.1Search83.379.7−3.7
MysteryMechanismScience51.849.1−2.7
AutomationBench-AA 1.0.6Agents74.271.8−2.4
LMCAGeneral QA70.668.3−2.3
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine43.941.7−2.2
DAXBench v2Code88.386.2−2.1
ARC-AGI-1 Semi-PrivateGeneral QA98.596.5−2.0
DTBenchGeneral QA99.297.3−1.9
AA-Briefcase v1.1 EloAgents17851816+1.7
τ³-Banking (Mercor run, bm25_grep retrieval)Agents55.754.0−1.7
HealthBench (length-adjusted, Artificial Analysis)Health & Medicine60.462.1+1.7
Excel Modeling Benchmark v1.1Finance & Accounting77.475.9−1.5
Vibe Code Bench v1.1Code93.892.4−1.4
LMArena Agent Arena Confirmed SuccessAgents22.019.5−1.2
DualEntry Accounting AI BenchmarkFinance & Accounting81.280.2−1.0
LMArena Agent Arena Bash ReliabilityAgents14.412.7−0.9
HealthBench ProfessionalHealth & Medicine71.470.8−0.6
GDPval-AA v2.1Professional Work18311838+0.3
OTIS Mock AIME 2024-2025Math99.7100.0+0.3
LMArena Agent Arena Tool HallucinationAgents1.00.9−0.1
ProofBench v1.1Math100.0100.00.0

Claude Fable 5.2vsClaude Sonnet 5.5normalised score marginsData Nov 15, 2026Data: Artificial Analysis, Epoch AI, Arena and 22 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…