Atlas

Compare Models

Claude Fable 5.2 vs Opus 4.8

Claude Fable 5.2 leads by 10.5 capability points (≈95% interval +7.9 to +13.1): a clear lead over Opus 4.8 · P(higher) ≈ 100%

Wins 49 of 49 shared benchmarks (0 lost)

Capability vs. relative cost
Claude Fable 5.2Opus 4.8Closed 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 5.2vsOpus 4.8capability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins49 – 0 · 49 shared

Comparing matching setups only (2 benchmarks excluded) · suites counted once
← Claude Fable 5.2Opus 4.8 →

Claude Fable 5.2vsOpus 4.8wins by categoryData Nov 15, 2026Data: Epoch AI, Artificial Analysis, Anthropic and 18 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Opus 4.8 minus Claude Fable 5.2.

ARC-AGI-3 (Semi-Private)General QA97.31.5−95.7
Clocktower BenchGames17591195−80.9
Humanity's Last Exam DiamondGeneral QA80.223.1−57.1
Mystery Game PuzzlesGames90.036.0−54.0
Furniture Assembly BenchmarkMultimodal90.042.5−47.5
FrontierMath Tier 4 (v2) - PrivateMath100.056.1−43.9
SimpleQA Verified (Epoch anti-abstention)General QA94.053.0−41.0
Bug Hunt BenchCode55.214.3−41.0
FrontierCode 2.0 - Pass RateCode54.716.7−38.0
LMArena Code Arena WebDevCode18361555−36.7
GameBench 2Games99.162.5−36.7
SWE-bench MultimodalCode74.138.4−35.7
FrontierCode 2.0 - ScoreCode50.214.9−35.3
Remote Labor Index (RLI)Professional Work41.38.3−33.0
HieroglyphBenchMultimodal43.614.0−29.6
AutomationBench-AA 1.0.6Agents74.245.6−28.6
AutomationBench 1.0.6 (Zapier)Agents44.316.9−27.4
Chess Puzzles (task version 2.0)Games79.052.0−27.0
AA-Briefcase v1.1 EloAgents17851320−25.5
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine43.919.4−24.4
SWE-Bench Pro (Public Dataset)Code93.069.2−23.8
HardGeoBench <=100miMultimodal60.036.3−23.7
Humanity's Last ExamGeneral QA73.249.8−23.4
APEX-Agents 1.1Professional Work87.664.4−23.2
ARC-AGI-2General QA93.872.1−21.7
HardGeoBench Avg DistanceMultimodal5521245−21.4
HardGeoBench <=10miMultimodal48.827.5−21.2
DeepSWE v1.1Code74.354.4−19.9
BuseyBench SVGMultimodal7.25.2−19.6
Chartography (with tools)Multimodal91.675.0−16.6
GDPval-AA v2.1Professional Work18311461−16.3
HardGeoBench Country AccuracyMultimodal70.053.8−16.2
Conceptual Reasoning IndexIndexes79.968.3−15.9
FrontierMath Tiers 1-3 (v2) - PrivateMath95.180.0−15.1
Vals RSI Index v1.1 (four tasks)Agents37.924.0−13.9
SciCode 1.0.1 (Artificial Analysis)Code67.754.4−13.3
LMCAGeneral QA70.657.5−13.1
ACCoRDGeneral QA89.477.4−12.0
Epoch Capabilities Index (ECI)Indexes174160−11.6
HealthBench ProfessionalHealth & Medicine71.460.3−11.1
Vibe Code Bench v1.1Code93.882.7−11.0
MineBench — Bradley-Terry RatingGames19321786−8.8
ARC-AGI-1 Semi-PrivateGeneral QA98.592.5−6.0
OfficeQAProfessional Work83.477.6−5.8
GPQA DiamondGeneral QA93.988.4−5.5
DTBenchGeneral QA99.294.9−4.3
Tax Agent BenchFinance & Accounting67.064.9−2.1
OTIS Mock AIME 2024-2025Math99.798.3−1.4
HealthBenchHealth & Medicine60.158.8−1.3

Claude Fable 5.2vsOpus 4.8normalised score marginsData Nov 15, 2026Data: Epoch AI, Artificial Analysis, Anthropic and 18 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…