Atlas

Compare Models

Claude Fable 5.2 vs GPT-5.5

Claude Fable 5.2 leads by 10.1 capability points (≈95% interval +7.4 to +12.7): a clear lead over GPT-5.5 · P(higher) ≈ 100%

Wins 41 of 44 shared benchmarks (2 lost, 1 tied)

Capability vs. relative cost
Claude Fable 5.2GPT-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 5.2vsGPT-5.5capability vs. relative costData Nov 17, 2026Village LLM Atlasatlas.village.dev

Benchmark wins41 – 2 · 44 shared, 1 tied

Comparing matching setups only (16 benchmarks excluded) · suites counted once
← Claude Fable 5.2GPT-5.5 →

Claude Fable 5.2vsGPT-5.5wins by categoryData Nov 17, 2026Data: Epoch AI, Artificial Analysis, Conceptual Reasoning Benchmark Authors and 16 othersVillage LLM Atlasatlas.village.dev

Margins

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

ARC-AGI-3 (Semi-Private)General QA97.30.4−96.8
Humanity's Last Exam DiamondGeneral QA80.227.9−52.3
Furniture Assembly BenchmarkMultimodal90.044.2−45.8
ChartographyMultimodal73.431.0−42.4
GameBench 2Games99.159.9−39.2
Clocktower BenchGames17591486−39.2
SWE-Bench Pro (Public Dataset)Code93.058.6−34.4
Mystery Game PuzzlesGames90.056.0−34.0
AA-Briefcase v1.1 EloAgents17371135−33.0
SimpleQA Verified (Epoch anti-abstention)General QA94.063.0−31.0
Humanity's Last ExamGeneral QA73.243.8−29.4
AutomationBench 1.0.6 (Zapier)Agents44.316.6−27.7
FrontierMath Tier 4 (v2) - PrivateMath100.072.5−27.5
AutomationBench-AA 1.0.6Agents74.247.3−26.9
Vibe Code Bench v1.1Code93.869.8−23.9
HieroglyphBenchMultimodal43.622.8−20.8
Blueprint-Bench 2Multimodal0.60.4−20.4
SWE-bench MultilingualCode96.677.7−18.9
OfficeQA ProProfessional Work72.954.1−18.8
GDPval-AA v2.1Professional Work17511357−17.4
APEX-Agents 1.1Professional Work87.670.3−17.3
Conceptual Reasoning IndexIndexes79.967.3−17.3
LMCAGeneral QA70.654.3−16.3
Chess Puzzles (task version 2.0)Games79.063.0−16.0
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine43.928.9−15.0
APEX-Accounting 1.1Finance & Accounting70.456.3−14.1
GBENCH One-shot Coding GScoreGames74.961.6−13.3
Vending-Bench 2Games96287524−12.7
BuseyBench SVGMultimodal7.26.0−11.7
SciCode 1.0.1 (Artificial Analysis)Code67.756.1−11.6
Epoch Capabilities Index (ECI)Indexes174160−10.6
CritPt (avg@5)Science35.725.4−10.3
DeepSWE v1.1Code74.364.4−9.9
FrontierMath Tiers 1-3 (v2) - PrivateMath95.185.3−9.8
ACCoRDGeneral QA89.480.5−8.9
ARC-AGI-2General QA93.885.0−8.7
Tax Agent BenchFinance & Accounting67.060.5−6.6
MineBench — Bradley-Terry RatingGames19321846−5.2
ARC-AGI-1 Semi-PrivateGeneral QA98.595.0−3.5
GPQA DiamondGeneral QA93.990.7−3.2
DTBenchGeneral QA99.296.0−3.2
AA-LCR v1.1Search83.384.3+1.0
OTIS Mock AIME 2024-2025Math99.7100.0+0.3
DAXBench v2Code88.388.30.0

Claude Fable 5.2vsGPT-5.5normalised score marginsData Nov 17, 2026Data: Epoch AI, Artificial Analysis, Conceptual Reasoning Benchmark Authors and 16 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…