Atlas

Compare Models

Claude Fable 5 vs DeepSeek V4.1 Flash 1017

Claude Fable 5 leads by 2.5 capability points (≈95% interval +0.0 to +5.0): a clear lead over DeepSeek V4.1 Flash 1017 · P(higher) ≈ 98%

Wins 41 of 52 shared benchmarks (11 lost)

Capability vs. relative cost
Claude Fable 5DeepSeek V4.1 Flash 1017Closed 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 5vsDeepSeek V4.1 Flash 1017capability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins41 – 11 · 52 shared

Comparing matching setups only (18 benchmarks excluded) · suites counted once
← Claude Fable 5DeepSeek V4.1 Flash 1017 →

Claude Fable 5vsDeepSeek V4.1 Flash 1017wins by categoryData Nov 15, 2026Data: Arena, Epoch AI, Mercor and 23 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is DeepSeek V4.1 Flash 1017 minus Claude Fable 5.

FrontierMath Tier 4 (v2) - PrivateMath90.246.3−43.9
LisanBench Difficulty-WeightedGames45627733+34.0
Humanity's Last Exam DiamondGeneral QA45.620.8−24.8
LisanBench Path LengthGames1368021107+23.1
NoCoT-Bench NCRI 15.2General QA11981.6−20.6
NoCoT-Bench NCKI (kspine v3)General QA128102−18.4
ALE-BenchCode24632012−16.0
FrontierMath Tiers 1-3 (v2) - PrivateMath87.071.2−15.8
Humanity's Last ExamGeneral QA57.842.6−15.2
Terminal-Bench 4.0Code44.531.0−13.5
Excel Modeling BenchmarkFinance & Accounting73.761.8−11.9
EnterpriseBench: CoreCraftProfessional Work70.358.5−11.8
GBENCH One-shot Coding GScoreGames65.653.8−11.8
ARC-AGI-2General QA89.278.1−11.1
ReactBench (39 tasks)Code44.634.4−10.3
LMArena Agent Arena Net Improvement (Steering Burden)Agents7.32.0−10.0
Concept Synth INDUCTION Challenge100 Round 1General QA31.021.0−10.0
APEX-SWECode58.949.1−9.8
Bug Hunt BenchCode27.637.3+9.7
Concept Synth INDUCTION Challenge64 Round 1General QA42.232.8−9.4
Terminal-Bench 4.0 (Artificial Analysis)Code42.433.8−8.6
CharXiv Reasoning (with tools)Multimodal93.585.6−7.9
DeepSWE v1.1Code69.977.0+7.1
MMMU ProMultimodal72.779.7+7.0
AA-Briefcase v1.1 EloAgents15361416−6.6
LMArena Agent Arena Praise vs ComplaintAgents14.31.2−6.5
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine20.626.7+6.1
DAYJOB: Finance (October 2026 rescoring)Finance & Accounting18.312.5−5.8
GameBench 2Games75.269.9−5.3
Ruby LLM Benchmarks — Program FixerCode77.272.0−5.3
DAYJOB: Healthcare (October 2026 rescoring)Health & Medicine11.26.0−5.2
Tax Agent BenchFinance & Accounting69.874.9+5.1
LMArena Code Arena WebDevCode16251664+5.0
APEX-Agents 1.1Professional Work74.579.4+4.9
LiveBench 2.0Agents83.679.2−4.4
Humanity's Sixth SenseMultimodal34.530.1−4.4
LMArena Agent Arena Bash ReliabilityAgents10.42.4−4.0
ARC-AGI-1 Semi-PrivateGeneral QA98.594.5−4.0
ZeroBench Main Mean pass@1 (five samples)Multimodal15.811.8−4.0
Mystery Game PuzzlesGames52.049.0−3.0
Vibe Code Bench v1.1Code90.487.7−2.6
RedlineBenchLegal42.444.9+2.5
Arena Image-to-WebDevCode16211609−2.5
APEX-1 (Mercor, mean score)Professional Work66.068.3+2.3
Chartography (with tools)Multimodal85.283.0−2.2
LMArena Agent Arena Steering BurdenAgents7.94.0−1.9
APEX-Accounting 1.1Finance & Accounting61.059.3−1.7
LMArena Agent Arena Confirmed SuccessAgents5.12.3−1.4
Chess Puzzles (task version 2.0)Games47.046.0−1.0
ZeroBench (Main questions, pass@5)Multimodal24.023.0−1.0
OTIS Mock AIME 2024-2025Math100.099.4−0.6
LMArena Agent Arena Tool HallucinationAgents0.40.3−0.1

Claude Fable 5vsDeepSeek V4.1 Flash 1017normalised score marginsData Nov 15, 2026Data: Arena, Epoch AI, Mercor and 23 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…