Atlas

Compare Models

Claude Fable 5 vs Kimi K3.1 Pro

Kimi K3.1 Pro leads by 0.4 capability points (≈95% interval −2.0 to +2.8): statistically tied · P(higher) ≈ 62%

Claude Fable 5 wins 28 of 49 shared benchmarks (20 lost, 1 tied)

Capability vs. relative cost
Claude Fable 5Kimi K3.1 ProClosed 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 5vsKimi K3.1 Procapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins28 – 20 · 49 shared, 1 tied

Comparing matching setups only (26 benchmarks excluded) · suites counted once
← Claude Fable 5Kimi K3.1 Pro →

Claude Fable 5vsKimi K3.1 Prowins by categoryData Nov 15, 2026Data: Arena, Epoch AI, Moonshot AI and 19 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Kimi K3.1 Pro minus Claude Fable 5.

BenchCAD - geometric overlapMultimodal38.492.7+54.3
FrontierMath Tier 4 (v2) - PrivateMath90.265.9−24.3
Terminal-Bench 4.0Code44.565.7+21.2
Furniture Assembly BenchmarkMultimodal35.855.8+20.0
Terminal-Bench-Science 0.1Science21.438.1+16.7
ZeroBench (Main questions, pass@5)Multimodal24.039.0+15.0
ZeroBench Main Mean pass@1 (five samples)Multimodal15.829.8+14.0
ALE-BenchCode24632132−11.7
MMMU ProMultimodal72.784.1+11.4
Mystery Game PuzzlesGames52.042.0−10.0
NoCoT-Bench NCRI 15.2General QA119103−8.9
GameBench 2Games75.284.0+8.8
FrontierMath Tiers 1-3 (v2) - PrivateMath87.078.2−8.8
NoCoT-Bench NCKI (kspine v3)General QA128116−8.2
LMArena Code Arena WebDevCode16251688+8.2
AA-Briefcase v1.1 EloAgents15361678+7.8
ACCoRDGeneral QA85.077.5−7.5
MineBench — Bradley-Terry RatingGames18691992+7.5
GBENCH One-shot Coding GScoreGames65.658.2−7.4
LiveBench 2.0Agents83.676.7−6.9
Humanity's Last ExamGeneral QA57.851.1−6.7
Bug Hunt BenchCode27.634.3+6.7
Conceptual Reasoning IndexIndexes72.767.8−6.6
Humanity's Last Exam DiamondGeneral QA45.639.4−6.2
Tax Agent BenchFinance & Accounting69.863.9−6.0
Vibe Code Bench v1.1Code90.484.9−5.5
DTBenchGeneral QA98.493.3−5.1
Chartography (with tools)Multimodal85.290.2+5.0
Concept Synth INDUCTION Challenge64 Round 1General QA42.246.9+4.7
LMArena Agent Arena Net Improvement (Steering Burden)Agents7.34.9−4.6
EnterpriseBench: CoreCraftProfessional Work70.366.1−4.2
ProofBench v1.1Math95.091.0−4.0
LMArena Text Arena EloChat & Writing15041483−3.6
LMCAGeneral QA61.157.6−3.5
LMArena Agent Arena Praise vs ComplaintAgents14.38.0−3.2
Concept Synth INDUCTION Challenge100 Round 1General QA31.034.0+3.0
Chess Puzzles (task version 2.0)Games47.045.0−2.0
LMArena Agent Arena Bash ReliabilityAgents10.47.2−1.6
APEX-Agents 1.1Professional Work74.575.9+1.4
APEX-Accounting 1.1Finance & Accounting61.062.4+1.4
DAYJOB: Finance (October 2026 rescoring)Finance & Accounting18.317.0−1.3
BrowseCompSearch93.392.1−1.2
LMArena Agent Arena Confirmed SuccessAgents5.12.7−1.2
LMArena Agent Arena Steering BurdenAgents7.96.7−0.6
Prophet Arena Arena Score (Fixed Context)Forecasting84.985.4+0.5
OfficeQA ProProfessional Work67.167.4+0.3
LMArena Agent Arena Tool HallucinationAgents0.41.0+0.3
CharXiv Reasoning (with tools)Multimodal93.593.4−0.1
OTIS Mock AIME 2024-2025Math100.0100.00.0

Claude Fable 5vsKimi K3.1 Pronormalised score marginsData Nov 15, 2026Data: Arena, Epoch AI, Moonshot AI and 19 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…