Atlas

Compare Models

Claude Fable 5.1 vs Claude Haiku 5.5

Claude Fable 5.1 leads by 4.4 capability points (≈95% interval +2.1 to +6.7): a clear lead over Claude Haiku 5.5 · P(higher) ≈ 100%

Wins 142 of 163 shared benchmarks (19 lost, 2 tied)

Capability vs. relative cost
Claude Fable 5.1Claude Haiku 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 5.1vsClaude Haiku 5.5capability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins142 – 19 · 163 shared, 2 tied

Comparing matching setups only (30 benchmarks excluded) · suites counted once
← Claude Fable 5.1Claude Haiku 5.5 →

Claude Fable 5.1vsClaude Haiku 5.5wins by categoryData Nov 15, 2026Data: Mercor, Artificial Analysis, Vals AI and 47 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Claude Haiku 5.5 minus Claude Fable 5.1.

KernelBench Mega Normalized (RTX PRO 6000)Code64.70.0−64.7
InferenceBenchCode9.34.4−58.2
Toloka Arena (v3)Agents71.021.0−50.0
SimpleQA Verified (Epoch anti-abstention)General QA70.823.8−47.0
FrontierMath Tier 4 (v2) - PrivateMath87.846.3−41.5
TaxCalcBench TY25Finance & Accounting42.02.0−40.0
Roboflow Vision Evals — Reasoning, Strict (September)Multimodal36.272.8+36.6
NoCoT-Bench NCRI 15.2General QA12868.3−32.8
AA-PersonalityBench Conviction indexChat & Writing10067.6−32.7
ALE-BenchCode28491996−30.2
NoCoT-Bench NCKI (kspine v3)General QA12885.9−29.4
ARC-AGI-2General QA90.061.3−28.7
Mystery Game PuzzlesGames58.030.0−28.0
LMArena Agent Arena Net Improvement (Steering Burden)Agents14.80.5−27.1
Artificial Analysis Coding Agent Index v1.5Indexes62.236.8−25.4
AA-PersonalityBench Warmth indexChat & Writing94.369.5−25.0
BuseyBench SVGMultimodal7.04.6−24.7
Terminal-Bench 4.0Code57.933.3−24.6
Roboflow Vision Evals — Counting, Strict (September)Multimodal48.673.0+24.4
Extended NYT Connections 940 (Quadratic v1)Games90.065.7−24.3
AA-PersonalityBench Imagination indexChat & Writing86.763.8−23.1
Furniture Assembly BenchmarkMultimodal70.047.5−22.5
BrowseComp (Mercor 130-question subset)Search78.356.3−22.0
AA-PersonalityBench Informativeness indexChat & Writing96.275.3−21.2
AA-PersonalityBench Expressiveness indexChat & Writing98.177.2−21.2
AA-PersonalityBench Transparency indexChat & Writing98.177.2−21.2
MineBench — Bradley-Terry RatingGames19281582−21.0
DeepResearch Bench II (Mercor 130-task set)Search55.034.3−20.7
Bug Hunt BenchCode41.020.5−20.5
AA-3D v1.0 EloCode1139998−20.5
BullshitBench v2 LeaderboardSafety77.057.0−20.0
LisanBench Difficulty-WeightedGames45732711−19.9
LMArena Code Arena WebDevCode17421594−19.4
AA-PersonalityBench Organization indexChat & Writing60.041.0−19.2
GameBench 2Games94.075.1−19.0
ChessBench CoherenceGames61.980.7+18.8
Terminal-Bench 2.1 Extended (Mercor held-out)Code19.537.7+18.2
Excel Modeling BenchmarkFinance & Accounting76.759.0−17.7
MedXpertQA MM Extended (Mercor held-out)Health & Medicine66.048.4−17.6
LMArena Agent Arena Praise vs ComplaintAgents31.6-3.5−17.6
MedXpertQA MM (200-question set)Health & Medicine83.767.8−15.9
MMMU-Pro Extended (Mercor held-out)Multimodal47.631.9−15.7
Tax Agent BenchFinance & Accounting77.962.2−15.7
FrontierMath Tiers 1-3 (v2) - PrivateMath90.275.1−15.1
Finance Agent v2 - General Qualitative AnalysisFinance & Accounting77.061.9−15.1
Eval Connections One-Shot with TrapsGames82.067.0−15.0
HardGeoBench <=10miMultimodal45.030.0−15.0
SM BenchSafety89.074.3−14.7
SWE Atlas v1.1 - OverallCode74.159.4−14.7
HardGeoBench Avg DistanceMultimodal8451319−14.6
Agent Initiative BenchmarkAgents7.56.0−14.5
LMCAGeneral QA65.551.1−14.4
WeirdML v3Science26.011.7−14.3
LMArena Text Arena EloChat & Writing15031422−14.0
Roboflow Vision Evals — Detection mAP@75 (September)Multimodal40.454.3+13.9
GBENCH One-shot Coding GScoreGames69.355.6−13.7
Immigration BenchLegal75.461.8−13.7
Roboflow Vision Evals — Identification, Strict (September)Multimodal93.880.2−13.6
AA-PersonalityBench Slop indexChat & Writing23.810.5−13.5
Roboflow Vision Evals — Detection mAP@50:95 (September)Multimodal39.653.0+13.4
EBR-bench v4 (card ban, single agent)Games57.143.8−13.3
Conceptual Reasoning IndexIndexes74.965.3−13.3
ChessBench EloGames1095875−13.1
SUPERChemScience74.762.0−12.7
MysteryMechanismScience47.735.1−12.6
Harvey LAB (Mercor run, undisclosed ~213-task cohort)Legal17.44.9−12.5
LiveBench 2.0Agents88.075.6−12.3
Humanity's Last Exam (Text-Only)General QA54.642.7−11.9
Roboflow Vision Evals — Identification, Judged (September)Multimodal97.986.5−11.4
Excel Modeling Benchmark v1.1Finance & Accounting76.765.5−11.3
ProofBench v1.1Math100.089.0−11.0
τ³-Banking (Mercor run, bm25_grep retrieval)Agents54.043.0−11.0
Code2Video-AA EloCode13251192−10.8
AutomationBench 1.0.6 (Zapier)Agents22.412.0−10.4
CritPt v2 (agentic, avg@5)Science76.666.3−10.3
KernelBench CUDA Normalized (RTX PRO 6000)Code51.641.4−10.2
DAXBench v2Code91.581.9−9.6
AA-LCR Extended (Mercor held-out)Search73.063.8−9.2
AA-LCR v1.1Search87.578.3−9.2
APEX-1 (Mercor, mean score)Professional Work73.864.7−9.1
Blueprint-Bench 2Multimodal0.40.3−8.9
DualEntry Accounting AI BenchmarkFinance & Accounting85.176.2−8.9
APEX-Agents 1.1Professional Work77.168.3−8.8
HardGeoBench <=100miMultimodal53.845.0−8.7
CritPt (avg@5)Science31.122.6−8.6
APEX-SWECode63.655.1−8.5
ARC-AGI-1 Semi-PrivateGeneral QA97.589.0−8.5
Design Arena Overall Code CategoriesCode13311280−8.5
ProgramBench (Mercor 100-task subset)Code11.33.0−8.3
SciCode 1.0.1 (Artificial Analysis)Code63.155.0−8.1
SWE-bench Verified Extended (Mercor held-out)Code61.353.3−8.0
Medical Long Context Reasoning (MLCR-AA)Health & Medicine71.163.3−7.8
SWE-bench VerifiedCode96.688.9−7.7
HardGeoBench Country AccuracyMultimodal58.851.3−7.5
DeepSWE v1.1Code67.359.9−7.4
Ramp Accounting BenchFinance & Accounting49.742.3−7.4
FrontierPhysics v0.1Science11.34.2−7.1
Chess Puzzles (task version 2.0)Games48.055.0+7.0
MMMU ProMultimodal86.179.1−7.0
FrontierML ($200 wallet)Agents41.934.9−7.0
LisanBench Path LengthGames1510012897−6.8
Finance Agent v2 - Market AnalysisFinance & Accounting69.563.1−6.5
LMArena Agent Arena Confirmed SuccessAgents14.41.6−6.4
LMArena Agent Arena Steering BurdenAgents14.82.4−6.2
Finance Agent v2 - Disclosure AnalysisFinance & Accounting70.364.3−6.0
SWE-bench MultilingualCode91.185.1−6.0
GDPval-AA v2.1Professional Work17651629−6.0
Harvey LAB-AA v1.1 (hallucination-gated)Legal6.40.6−5.8
CVDP (517-problem set)Code62.657.0−5.6
DeepFinanceFinance & Accounting45.640.0−5.6
MMLU ProGeneral QA93.087.4−5.6
Epoch Capabilities Index (ECI)Indexes167160−5.5
SWE Atlas - Codebase QnACode43.348.7+5.4
Finance Agent v2 - General Quantitative AnalysisFinance & Accounting79.774.4−5.4
AA-Briefcase v1.1 EloAgents16711573−5.4
Finance Agent v2 - PrecedentsFinance & Accounting32.227.0−5.2
Roboflow Vision Evals (October 2026, five tasks)Multimodal74.769.6−5.1
Finance Agent v2 - All PassFinance & Accounting47.842.9−4.9
Finance Agent v2 - AdjustmentsFinance & Accounting54.449.6−4.8
ChessBench AccuracyGames71.366.9−4.7
APEX-Accounting 1.1Finance & Accounting66.762.2−4.5
FrontierCode 2.0 - Pass RateCode36.431.9−4.5
HealthBench ProfessionalHealth & Medicine69.264.8−4.5
OSWorld 2.1 offline - reported partial reward, step limit unspecifiedAgents76.872.4−4.4
SWE-Marathon v1.1Code45.650.0+4.4
FORTRESS Average Risk Score (ARS)Safety13.79.5+4.2
FrontierCode 2.0 - ScoreCode33.028.9−4.0
ChessBenchGames45.341.3−4.0
AA-AnalystAgent (pass^5)Professional Work57.553.8−3.7
FrontierCode 1.1 Extended - ScoreCode62.058.4−3.6
GDPxProfessional Work8.54.9−3.6
AdvancedIFChat & Writing84.280.8−3.4
CursorBench 4.0Code51.848.4−3.4
Terminal-Bench 2.1Code81.685.0+3.4
Harvey LAB-AA v1.0 (criterion pass rate)Legal93.389.9−3.4
Roboflow Vision Evals — Detection mAP@50 (September)Multimodal65.068.2+3.2
FrontierCode 1.1 Extended - Pass RateCode67.964.7−3.2
Finance Agent v2 - Financial ModelingFinance & Accounting25.328.3+3.0
SAGEMath48.545.8−2.7
GDPval Extended (Mercor held-out rubric score)Professional Work68.766.0−2.7
HealthBenchHealth & Medicine63.961.2−2.7
DTBenchGeneral QA97.695.0−2.6
ACCoRDGeneral QA83.781.5−2.1
Finance Agent v2 - Earnings AnalysisFinance & Accounting75.573.5−2.0
MedCodeHealth & Medicine53.551.5−2.0
MedScribeHealth & Medicine91.389.5−1.8
Roboflow Vision Evals — Reasoning, Judged (September)Multimodal73.174.8+1.7
LMArena Agent Arena Tool HallucinationAgents0.8-2.1−1.4
LMArena Agent Arena Bash ReliabilityAgents10.17.4−1.3
HealthBench (length-adjusted, Artificial Analysis)Health & Medicine57.258.5+1.3
OTIS Mock AIME 2024-2025Math100.098.9−1.1
GDPval (Mercor 213-task rubric score)Professional Work86.685.5−1.1
CharXiv (descriptive and reasoning combined)Multimodal95.194.1−1.0
MazeBench ASCIIGames2.01.0−1.0
CharXiv Extended (Mercor held-out)Multimodal82.282.9+0.7
Finance Agent v2 - ComparablesFinance & Accounting46.045.3−0.7
Prophet Arena Arena Score (Fixed Context)Forecasting85.085.4+0.3
SpeciEval (Pooled English Runs)Safety87.286.9−0.3
ChartographyMultimodal46.246.4+0.2
Prophet Arena Arena Score (Agentic, OpenCode)Forecasting86.085.9−0.2
DistressBenchSafety71.571.7+0.1
AIME 2025Math100.0100.00.0
Roboflow Vision Evals — Counting, Judged (September)Multimodal73.073.00.0

Claude Fable 5.1vsClaude Haiku 5.5normalised score marginsData Nov 15, 2026Data: Mercor, Artificial Analysis, Vals AI and 47 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…