Atlas

Compare Models

Claude Fable 5 vs Gemini 4 Argon

Gemini 4 Argon leads by 2.4 capability points (≈95% interval −0.0 to +4.9): statistically tied · P(higher) ≈ 97%

Wins 73 of 106 shared benchmarks (30 lost, 3 tied)

Capability vs. relative cost
Claude Fable 5Gemini 4 ArgonClosed 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 5vsGemini 4 Argoncapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins30 – 73 · 106 shared, 3 tied

Comparing matching setups only (52 benchmarks excluded) · suites counted once
← Claude Fable 5Gemini 4 Argon →

Claude Fable 5vsGemini 4 Argonwins by categoryData Nov 15, 2026Data: Center for AI Safety, Epoch AI, Arena and 37 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Gemini 4 Argon minus Claude Fable 5.

VCT-Refusal (CAIS)Safety100.010.7−89.3
HPCT Refusal (CAIS)Safety100.029.0−71.0
HieroglyphBenchMultimodal22.969.7+46.8
ClockBench AccuracyMultimodal37.283.3+46.1
ChartographyMultimodal34.871.6+36.8
Furniture Assembly BenchmarkMultimodal35.872.5+36.7
AutomationBench 1.0.6 (Zapier)Agents17.150.1+33.0
MASKSafety92.861.0−31.8
Terminal-Bench 2.1Code88.057.7−30.3
TaxCalcBench TY25 (web search)Finance & Accounting44.074.0+30.0
Agent Red Teaming k=100 (CAIS)Safety39.410.2+29.1
BrowseComp Extended (Mercor held-out)Search44.572.0+27.5
ChessBench EloGames10461485+26.1
Concept Synth INDUCTION Challenge100 Round 1General QA31.057.0+26.0
Concept Synth INDUCTION Challenge64 Round 1General QA42.265.6+23.4
LLM Creative Story WritingChat & Writing2.90.8−23.2
TextQuests Game ProgressGames56.177.3+21.2
Humanity's Sixth SenseMultimodal34.555.4+20.9
SpatialViz-BenchMultimodal73.994.2+20.3
MineBench — Bradley-Terry RatingGames18692195+19.7
LisanBench Path LengthGames136807560−19.0
ERQAMultimodal67.886.5+18.8
NoCoT-Bench NCRI 15.2General QA119151+17.7
RiemannBenchMath60.076.0+16.0
Blueprint-Bench 2Multimodal0.40.5+15.8
MMMU ProMultimodal72.788.3+15.6
LMArena Agent Arena Bash RecoveryAgents9.6-21.4−15.5
ComplexConstraintsChat & Writing38.153.3+15.2
Machiavelli Deception (CAIS)Safety80.285.0−14.5
GameBench 2Games75.289.2+13.9
Agents' Last Exam (ALE)Agents25.739.5+13.8
MindCube TinyMultimodal83.196.7+13.7
GPQA DiamondGeneral QA83.396.4+13.1
APEX-Agents 1.1Professional Work74.587.4+12.9
Terminal-Bench 4.0Code44.557.4+12.9
BioTIER RefusalSafety100.087.2−12.8
Terminal-Bench 4.0 (Artificial Analysis)Code42.455.1+12.6
FrontierMath Tier 4 (v2) - PrivateMath90.278.0−12.2
Chess Puzzles (task version 2.0)Games47.059.0+12.0
Mystery Game PuzzlesGames52.064.0+12.0
Bug Hunt BenchCode27.638.7+11.1
GDPval Extended (Mercor held-out rubric score)Professional Work65.376.4+11.1
APEX-Accounting 1.1Finance & Accounting61.071.5+10.5
ALE-BenchCode24632747+10.1
ChessBench CoherenceGames82.692.6+10.0
BullshitBench v2 LeaderboardSafety56.066.0+10.0
MedXpertQA MM Extended (Mercor held-out)Health & Medicine56.966.7+9.8
DAYJOB: FinanceFinance & Accounting9.018.8+9.8
APEX-SWECode58.949.2−9.7
Tax Agent BenchFinance & Accounting69.879.0+9.2
PLE (Physical-Engineering's Last Exam)Professional Work34.943.2+8.3
TextQuests Average Cumulative HarmSafety23.726.1−8.2
DeepSWE v1.1Code69.977.9+8.0
LMArena Agent Arena Bash ReliabilityAgents10.4-5.4−7.9
ACCoRDGeneral QA85.077.3−7.7
LisanBench Difficulty-WeightedGames45623849−7.6
IntPhys 2Multimodal67.174.6+7.6
LMArena Code Arena WebDevCode16251682+7.4
Senior SWE-BenchCode34.727.7−7.0
Eval Connections One-Shot with TrapsGames85.092.0+7.0
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine20.613.9−6.7
Physical-Engineering's Last Exam (hidden test)Professional Work34.541.1+6.6
LMArena Agent Arena Net Improvement (Steering Burden)Agents7.310.7+6.5
Vals RSI Index v1.1 (four tasks)Agents24.330.6+6.2
Gray Swan IPI Attack Success (k=15)Safety6.50.7+5.8
HiL-Bench (Human-in-Loop Benchmark)Agents56.362.0+5.7
DAYJOB: HealthcareHealth & Medicine5.610.8+5.2
ChessBench AccuracyGames77.181.8+5.0
MirrorCode (ML, +Private, 2L)Code63.958.9−5.0
LMArena Agent Arena Praise vs ComplaintAgents14.324.1+4.9
FrontierCode 1.1 Extended - Pass RateCode70.966.1−4.7
WeirdML v3Science15.620.3+4.7
FrontierCode 1.1 Extended - ScoreCode64.960.6−4.4
LMArena Agent Arena Confirmed SuccessAgents5.113.2+4.1
ProofBench v1.1Math95.099.0+4.0
AA-LCR Extended (Mercor held-out)Search71.174.7+3.6
LMArena Text Arena EloChat & Writing15041522+3.2
Humanity's Last Exam DiamondGeneral QA45.642.5−3.1
AA-Briefcase v1.1 EloAgents15361485−2.8
MedCodeHealth & Medicine56.158.8+2.7
LMArena Agent Arena Steering BurdenAgents7.913.2+2.6
LMCAGeneral QA61.163.7+2.5
HardGeoBench Avg DistanceMultimodal800880−2.5
GBENCH One-shot Coding GScoreGames65.667.9+2.3
Humanity's Last ExamGeneral QA57.855.6−2.2
GDPval (Mercor 213-task rubric score)Professional Work83.885.9+2.1
ARC-AGI-2General QA89.291.3+2.1
SAGEMath51.953.6+1.8
Excel Modeling BenchmarkFinance & Accounting73.775.2+1.6
FrontierMath ErdősMath0.01.5+1.5
GDP.pdfProfessional Work29.828.4−1.4
HardGeoBench <=10miMultimodal41.342.5+1.3
HardGeoBench <=100miMultimodal56.355.0−1.2
HardGeoBench Country AccuracyMultimodal61.362.5+1.2
PCT Political Inconsistency (CAIS)Safety39.838.7+1.1
FrontierMath Tiers 1-3 (v2) - PrivateMath87.086.0−1.1
DTBenchGeneral QA98.497.6−0.8
LiveBench 2.0Agents83.684.1+0.5
Prophet Arena Arena Score (Fixed Context)Forecasting84.985.3+0.4
Conceptual Reasoning IndexIndexes72.772.4−0.3
LegalBenchLegal88.688.3−0.3
NoCoT-Bench NCKI (kspine v3)General QA128128+0.1
LMArena Agent Arena Tool HallucinationAgents0.40.50.0
AA-AnalystAgent (pass^5)Professional Work48.848.80.0
ARC-AGI-1 Semi-PrivateGeneral QA98.598.50.0
OTIS Mock AIME 2024-2025Math100.0100.00.0

Claude Fable 5vsGemini 4 Argonnormalised score marginsData Nov 15, 2026Data: Center for AI Safety, Epoch AI, Arena and 37 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…