Atlas

Compare Models

Claude Fable 5.2 vs Claude Opus 5

Claude Fable 5.2 leads by 6.2 capability points (≈95% interval +3.4 to +8.9): a clear lead over Claude Opus 5 · P(higher) ≈ 100%

Wins 62 of 66 shared benchmarks (4 lost)

Capability vs. relative cost
Claude Fable 5.2Claude Opus 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 Opus 5capability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins62 – 4 · 66 shared

Comparing matching setups only (5 benchmarks excluded) · suites counted once
← Claude Fable 5.2Claude Opus 5 →

Claude Fable 5.2vsClaude Opus 5wins by categoryData Nov 15, 2026Data: Artificial Analysis, Epoch AI, Anthropic and 19 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Claude Opus 5 minus Claude Fable 5.2.

Anthropic ECI (AECI), November 2026 fitIndexes177169−67.7
ARC-AGI-3 (Semi-Private)General QA97.330.2−67.1
EBR-bench v4 (card ban, single agent)Games90.545.7−44.8
LMArena Agent Arena Net Improvement (Steering Burden)Agents30.47.2−43.7
Chess Puzzles (task version 2.0)Games79.043.0−36.0
Humanity's Last Exam DiamondGeneral QA80.244.8−35.4
Mystery Game PuzzlesGames90.059.0−31.0
LMArena Agent Arena Praise vs ComplaintAgents73.412.2−30.6
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine43.913.9−30.0
Bug Hunt BenchCode55.225.7−29.5
Furniture Assembly BenchmarkMultimodal90.060.8−29.2
HieroglyphBenchMultimodal43.614.6−29.0
FrontierMath Tier 4 (v2) - PrivateMath100.073.2−26.8
FrontierCode 2.0 - Pass RateCode54.730.4−24.3
BioSecBench-Structure V1 (58 evaluations)Science43.720.7−23.0
FrontierCode 2.0 - ScoreCode50.227.4−22.8
CrossSearch (pass^3)Search56.034.0−22.0
HardGeoBench Avg DistanceMultimodal5521218−20.6
LMArena Code Arena WebDevCode18361691−19.0
FrontierSWE V2Code69.952.0−17.9
AutomationBench-AA 1.0.6Agents74.256.6−17.6
AutomationBench 1.0.6 (Zapier)Agents44.326.9−17.3
Humanity's Last ExamGeneral QA73.256.4−16.8
HardGeoBench <=10miMultimodal48.832.5−16.2
HardGeoBench Country AccuracyMultimodal70.053.8−16.2
CritPt v2 (agentic, avg@5)Science86.371.4−14.9
SWE-bench MultimodalCode74.159.4−14.7
MysteryMechanismScience51.837.4−14.4
GameBench 2Games99.184.8−14.3
LMArena Agent Arena Steering BurdenAgents33.04.4−14.3
SWE-Bench Pro (Public Dataset)Code93.079.2−13.8
HardGeoBench <=100miMultimodal60.046.3−13.7
τ³-Banking (Mercor run, bm25_grep retrieval)Agents55.744.0−11.7
SciCode 1.0.1 (Artificial Analysis)Code67.756.4−11.3
APEX-Agents 1.1Professional Work87.677.0−10.6
Tax Agent BenchFinance & Accounting67.076.7+9.6
FrontierMath Tiers 1-3 (v2) - PrivateMath95.185.6−9.5
DeepSWE v1.1Code74.365.0−9.3
BuseyBench SVGMultimodal7.26.3−8.7
Chartography (with tools)Multimodal91.683.0−8.6
Epoch Capabilities Index (ECI)Indexes174164−7.9
Conceptual Reasoning IndexIndexes79.974.7−7.2
AA-Briefcase v1.1 EloAgents17851654−7.1
HealthBenchHealth & Medicine60.167.1+7.0
CritPt (avg@5)Science35.729.1−6.6
LMCAGeneral QA70.664.5−6.2
APEX-Accounting 1.1Finance & Accounting70.464.3−6.1
MacroscopeBenchCode76.069.9−6.1
GPQA DiamondGeneral QA93.987.9−6.0
LMArena Agent Arena Confirmed SuccessAgents22.010.2−5.9
ACCoRDGeneral QA89.483.8−5.6
Vibe Code Bench v1.1Code93.888.4−5.4
Vals RSI Index v1.1 (four tasks)Agents37.933.0−4.9
MineBench — Bradley-Terry RatingGames19322008+4.6
GDPval-AA v2.1Professional Work18311729−4.5
BioMysteryBench (Vals Terminus 2)Science83.779.3−4.4
ARC-AGI-2General QA93.890.4−3.3
LMArena Agent Arena Bash ReliabilityAgents14.49.8−2.3
DAXBench v2Code88.386.2−2.1
HealthBench ProfessionalHealth & Medicine71.473.4+2.0
AA-LCR v1.1Search83.382.0−1.3
DTBenchGeneral QA99.297.9−1.3
ARC-AGI-1 Semi-PrivateGeneral QA98.597.5−1.0
ProofBench v1.1Math100.099.0−1.0
OTIS Mock AIME 2024-2025Math99.798.9−0.8
LMArena Agent Arena Tool HallucinationAgents1.00.90.0

Claude Fable 5.2vsClaude Opus 5normalised score marginsData Nov 15, 2026Data: Artificial Analysis, Epoch AI, Anthropic and 19 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…