Atlas

Compare Models

Claude Fable 5 vs Gemini 4 Carbon

Gemini 4 Carbon leads by 4.5 capability points (≈95% interval +1.8 to +7.1): a clear lead over Claude Fable 5 · P(higher) ≈ 100%

Wins 36 of 43 shared benchmarks (5 lost, 2 tied)

Capability vs. relative cost
Claude Fable 5Gemini 4 CarbonClosed weightsOpen weights

Benchmark wins5 – 36 · 43 shared, 2 tied

Comparing matching setups only (8 benchmarks excluded) · suites counted once
← Claude Fable 5Gemini 4 Carbon →
Data: Epoch AI, Arena, Surge AI and 14 othersatlas.village.dev

Margins

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

HieroglyphBenchPerception & media › Documents & charts22.976.0+53.1
Furniture Assembly BenchmarkPerception & media › Spatial & visual reasoning35.877.5+41.7
ChartographyPerception & media › Documents & charts34.872.5+37.7
Humanity's Sixth SensePerception & media › Spatial & visual reasoning34.571.1+36.6
AutomationBench 1.0.6 (Zapier)Tool & computer use › Customer service & operations17.153.0+35.9
Chess Puzzles (task version 2.0)Games47.078.0+31.0
FrontierCUA v1 pass@1Tool & computer use › Computer & browser use15.042.5+27.5
Bug Hunt BenchSoftware › Software engineering27.654.6+27.0
DAYJOB: Healthcare (October 2026 rescoring)Health & medicine11.235.2+24.0
Humanity's Last Exam DiamondKnowledge & reasoning › General knowledge45.669.1+23.5
Mystery Game PuzzlesGames52.073.0+21.0
Terminal-Bench 4.0 (Artificial Analysis)Software › Software engineering42.462.6+20.2
RiemannBenchMath60.080.0+20.0
Blueprint-Bench 2Perception & media › Spatial & visual reasoning0.40.6+20.0
LMArena Agent Arena Net Improvement (Steering Burden)Tool & computer use › General agents5.916.1+20.0
LMArena Code Arena WebDevSoftware › Frontend & design16261769+19.0
MMMU ProKnowledge & reasoning › General knowledge72.789.2+16.5
APEX-Agents 1.1Professional › General knowledge work74.590.4+15.9
LMArena Agent Arena Praise vs ComplaintTool & computer use › General agents11.539.4+14.0
GPQA DiamondScience › Multidisciplinary83.396.5+13.2
EnterpriseBench: CoreCraftTool & computer use › Customer service & operations70.380.5+10.2
APEX-Accounting 1.1Professional › Finance, accounting & tax61.070.7+9.7
LMArena Agent Arena Bash ReliabilityTool & computer use › General agents9.0-5.7−7.4
Tax Agent BenchProfessional › Finance, accounting & tax69.876.4+6.6
Gray Swan IPI Attack Success (k=15)Safety & alignment6.50.0+6.5
LMArena Agent Arena Confirmed SuccessTool & computer use › General agents4.417.2+6.4
LMArena Agent Arena Steering BurdenTool & computer use › General agents5.517.9+6.2
HiL-Bench (Human-in-Loop Benchmark)Tool & computer use › General agents56.350.3−6.0
DAYJOB: Finance (October 2026 rescoring)Professional › Finance, accounting & tax18.323.8+5.5
GBENCH One-shot Coding GScoreGames63.669.0+5.4
ARC-AGI-2Knowledge & reasoning › Abstract reasoning89.294.2+5.0
FrontierMath Tier 4 (v2) - PrivateMath90.295.1+4.9
LLM Debate BenchmarkLanguage & writing › Language & wordplay17051676−4.1
AA-Briefcase v1.1 EloProfessional › General knowledge work15361611+4.1
LMArena Text Arena EloLanguage & writing › General assistance15041526+3.8
FrontierMath Tiers 1-3 (v2) - PrivateMath87.090.5+3.5
Vals RSI Index v1.1 (four tasks)Software › Data & ML24.327.7+3.4
FrontierMath ErdősMath0.02.9+2.9
Vibe Code Bench v1.1Software › Frontend & design90.489.1−1.3
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & medicine20.619.4−1.1
LMArena Agent Arena Tool HallucinationTool & computer use › Tool calling0.50.8+0.2
ARC-AGI-1 Semi-PrivateKnowledge & reasoning › Abstract reasoning98.598.50.0
OTIS Mock AIME 2024-2025Math100.0100.00.0
Data: Epoch AI, Arena, Surge AI and 14 othersatlas.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…