Atlas

Compare Models

Claude Fable 5 vs Claude Opus 5.5

Claude Opus 5.5 leads by 4.3 capability points (≈95% interval +1.7 to +6.8): a clear lead over Claude Fable 5 · P(higher) ≈ 100%

Wins 137 of 162 shared benchmarks (19 lost, 6 tied)

Capability vs. relative cost
Claude Fable 5Claude Opus 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 5vsClaude Opus 5.5capability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins19 – 137 · 162 shared, 6 tied

Comparing matching setups only (60 benchmarks excluded) · suites counted once
← Claude Fable 5Claude Opus 5.5 →

Claude Fable 5vsClaude Opus 5.5wins by categoryData Nov 15, 2026Data: Mercor, Center for AI Safety, Arena and 55 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Claude Opus 5.5 minus Claude Fable 5.

BenchCAD - geometric overlapMultimodal38.496.2+57.8
Anthropic ECI (AECI), November 2026 fitIndexes169175+53.5
BrokenArXiv 06/2026Math47.8100.0+52.2
Olam Social Poker EloGames15531506−51.1
BrokenArXiv 05/2026Math44.595.0+50.5
Furniture Assembly BenchmarkMultimodal35.883.3+47.5
SWE Atlas Codebase QnA Extended (Mercor held-out)Code0.043.3+43.3
ClockBench AccuracyMultimodal37.277.2+40.0
InferenceBenchCode8.712.1+39.9
Terminal-Bench-Science 0.1Science21.458.7+37.3
VoxelBench Text RatingGames21012613+36.5
Terminal-Bench Science 0.1 (Vals)Science12.948.6+35.7
ActiveVisionMultimodal7.141.2+34.1
Concept Synth INDUCTION Challenge100 Round 1General QA31.065.0+34.0
LisanBench Path LengthGames1368024029+32.2
YC-BenchGames19775732873725+31.3
Terminal-Bench 2.1 Extended (Mercor held-out)Code16.847.5+30.7
KernelBench Mega Normalized (RTX PRO 6000)Code69.4100.0+30.6
Concept Synth INDUCTION Challenge64 Round 1General QA42.271.9+29.7
LMArena Code Arena WebDevCode16251814+24.7
KernelBench CUDA Normalized (RTX PRO 6000)Code51.676.3+24.6
Into the Unknown Integrity Node Pass RateScience66.590.4+23.9
LMArena Agent Arena Net Improvement (Steering Burden)Agents7.319.5+23.1
AutomationBench - release-reported metric and version unspecifiedAgents17.140.0+22.9
Terminal-Bench 4.0Code44.564.8+20.3
CadQueryEval (Corrected Geometry)Code80.0100.0+20.0
Chess Puzzles (task version 2.0)Games47.067.0+20.0
ProgramBench (Mercor 100-task subset)Code0.020.0+20.0
GameBench 2Games75.294.8+19.5
ZeroBench Main Mean pass@1 (five samples)Multimodal15.835.0+19.2
DAYJOB: HealthcareHealth & Medicine5.624.7+19.1
GlobeBenchGeneral QA7.326.3+19.1
Mystery Game PuzzlesGames52.071.0+19.0
ZeroBench (Main questions, pass@5)Multimodal24.043.0+19.0
ChessBench EloGames10461349+18.0
AdvancedIFChat & Writing65.282.8+17.6
Terminal-Bench 4.0 (Artificial Analysis)Code42.459.6+17.2
MineBench — Bradley-Terry RatingGames18692134+16.1
BrowseComp Extended (Mercor held-out)Search44.560.5+16.0
ZeroBench Main Questions pass^5Multimodal9.025.0+16.0
GeneBench-Pro Public Release SubsetScience14.030.0+16.0
SWE Atlas - Codebase QnACode33.148.9+15.8
SpatialViz-BenchMultimodal73.989.7+15.8
PLE (Physical-Engineering's Last Exam)Professional Work34.950.6+15.7
WeirdML v3Science15.631.2+15.6
SUPERChemScience64.479.6+15.2
RuneBenchGames5.76.7+15.1
LisanBench Difficulty-WeightedGames45625963+15.0
MirrorCode (ML, +Private, 2L)Code63.978.9+15.0
Agent Red Teaming k=100 (CAIS)Safety39.424.4+14.9
DAYJOB: FinanceFinance & Accounting9.023.9+14.9
AA-Briefcase v1.1 EloAgents15361801+14.5
HieroglyphBenchMultimodal22.937.4+14.5
MedXpertQA MM Extended (Mercor held-out)Health & Medicine56.971.2+14.3
MMLU ProGeneral QA79.493.4+14.0
MMMU ProMultimodal72.786.6+13.9
Physical-Engineering's Last Exam (hidden test)Professional Work34.548.3+13.8
MMMU-Pro Extended (Mercor held-out)Multimodal37.651.0+13.4
LMArena Agent Arena Praise vs ComplaintAgents14.340.3+13.0
Vals RSI Index v1.1 (four tasks)Agents24.337.3+13.0
Drone-BenchAgents84.397.3+13.0
MaxxBenchCode63.576.4+12.9
SWE-bench Verified Extended (Mercor held-out)Code78.766.0−12.7
Vals RSI Index v1.1 (historical five tasks)Agents24.537.1+12.7
Blueprint-Bench 2Multimodal0.40.5+12.6
Agents' Last Exam (ALE)Agents25.738.2+12.5
Bug Hunt BenchCode27.639.7+12.1
TaxCalcBench TY25Finance & Accounting30.042.0+12.0
Search ArenaSearch12281213−11.9
ComplexConstraintsChat & Writing38.149.9+11.8
Humanity's Last Exam RMS Calibration Error (CAIS)General QA25.513.8+11.7
PCT Political Inconsistency (CAIS)Safety39.829.1+10.7
GDPval Extended (Mercor held-out rubric score)Professional Work65.375.8+10.5
Roboflow Vision Evals (October 2026, five tasks)Multimodal71.782.0+10.3
Humanity's Sixth SenseMultimodal34.544.6+10.1
ALE-BenchCode24632748+10.1
MindCube TinyMultimodal83.193.1+10.0
LLM Creative Story WritingChat & Writing2.93.7+9.7
ReactBench (39 tasks)Code44.654.4+9.7
VCT-Refusal (CAIS)Safety100.090.4−9.6
RiemannBenchMath60.069.6+9.6
MedXpertQA MM (200-question set)Health & Medicine76.285.7+9.5
Humanity's Last Exam DiamondGeneral QA45.654.6+9.0
Into the Unknown Direction Node Pass RateScience71.080.0+9.0
APEX-SWECode58.967.6+8.7
GBENCH One-shot Coding GScoreGames65.673.9+8.3
Machiavelli Deception (CAIS)Safety80.277.5+8.2
LMArena Agent Arena Confirmed SuccessAgents5.120.7+7.8
AA-AnalystAgent (pass^5)Professional Work48.856.3+7.5
HardGeoBench Country AccuracyMultimodal61.368.8+7.5
ERQAMultimodal67.875.1+7.3
CharXiv (descriptive and reasoning combined)Multimodal89.196.4+7.3
Conceptual Reasoning IndexIndexes72.777.9+7.2
LMCAGeneral QA61.168.2+7.1
BullshitBench v2 LeaderboardSafety56.063.0+7.0
APEX-Agents 1.1Professional Work74.581.3+6.8
Roboflow Vision Evals (September 2026 protocol)Multimodal78.785.5+6.8
MedCodeHealth & Medicine56.149.8−6.3
HardGeoBench <=100miMultimodal56.362.5+6.3
SAGEMath51.945.8−6.1
Humanity's Last Exam (Text-Only)General QA51.557.5+6.0
BrowseComp (Mercor 130-question subset)Search82.588.5+6.0
ArXivMath 05/2026Math87.581.7−5.8
SciCode (main problems)Science21.527.3+5.8
APEX-AccountingFinance & Accounting56.461.8+5.4
LMArena Agent Arena Steering BurdenAgents7.918.6+5.3
ProofBench v1.1Math95.0100.0+5.0
FrontierMath Tier 4 (v2) - PrivateMath90.295.0+4.8
Senior SWE-BenchCode34.739.3+4.6
GDPval (Mercor 213-task rubric score)Professional Work83.888.3+4.5
Terminal-Bench 2.1Code88.083.5−4.5
HardGeoBench Avg DistanceMultimodal800655+4.5
DeepSWE v1.1Code69.974.2+4.3
FrontierMath Tiers 1-3 (v2) - PrivateMath87.091.2+4.2
Extended NYT Connections 940 (Quadratic v1)Games92.788.5−4.2
ARC-AGI-2General QA89.293.3+4.2
EnterpriseBench: CoreCraftProfessional Work70.374.4+4.1
CharXiv Extended (Mercor held-out)Multimodal80.784.7+4.0
FrontierCUA pass@1Agents40.036.0−4.0
Chartography (with tools)Multimodal85.289.0+3.8
LLM Debate BenchmarkChat & Writing17271701−3.6
ACCoRDGeneral QA85.088.4+3.5
MASKSafety92.889.4−3.4
GameDevBenchGames67.370.6+3.3
Humanity's Last Exam (with tools)General QA64.567.7+3.2
AA-LCR Extended (Mercor held-out)Search71.174.2+3.1
FrontierMath ErdősMath0.02.9+2.9
LiveBench 2.0Agents83.686.5+2.9
LMArena Agent Arena Bash RecoveryAgents9.615.0+2.7
SWE-bench VerifiedCode95.598.2+2.7
HiL-Bench (Human-in-Loop Benchmark)Agents56.359.0+2.7
ChessBench CoherenceGames82.680.1−2.5
HardGeoBench <=10miMultimodal41.343.8+2.5
ChessBench AccuracyGames77.179.3+2.4
LMArena Agent Arena Bash ReliabilityAgents10.415.1+2.3
Harvey LAB-AA v1.0 (criterion pass rate)Legal93.691.2−2.3
Excel Modeling BenchmarkFinance & Accounting73.775.9+2.3
Medical Long Context Reasoning (MLCR-AA)Health & Medicine64.466.7+2.2
ArXivMath 06/2026Math85.483.3−2.1
Clocktower BenchGames15341547+1.9
TextQuests Average Cumulative HarmSafety23.724.2−1.6
BioTIER RefusalSafety100.098.7−1.3
TextQuests Game ProgressGames56.157.3+1.2
IntPhys 2Multimodal67.168.3+1.2
Eval Connections One-Shot with TrapsGames85.086.0+1.0
FrontierCode 1.1 Extended - ScoreCode64.365.2+1.0
FrontierCode 1.1 Extended - Pass RateCode69.870.7+0.9
SWE-bench MultilingualCode94.895.5+0.7
ObviousBenchGeneral QA99.3100.0+0.7
Prophet Arena Arena Score (Fixed Context)Forecasting84.985.5+0.6
DTBenchGeneral QA98.498.9+0.5
RedlineBenchLegal42.442.0−0.4
LMArena Text Arena EloChat & Writing15041506+0.4
Ruby LLM Benchmarks — Program FixerCode77.277.5+0.3
APEX-1 (Mercor, mean score)Professional Work66.065.8−0.2
LMArena Agent Arena Tool HallucinationAgents0.40.7+0.1
AIME 2025Math100.0100.00.0
ARC-AGI-1 Semi-PrivateGeneral QA98.598.50.0
HPCT Refusal (CAIS)Safety100.0100.00.0
Into the UnknownScience0.00.00.0
OTIS Mock AIME 2024-2025Math100.0100.00.0
SWE-TogetherCode68.868.80.0

Claude Fable 5vsClaude Opus 5.5normalised score marginsData Nov 15, 2026Data: Mercor, Center for AI Safety, Arena and 55 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…