Atlas

Compare Models

Claude Fable 5 vs Grok 4.7

Claude Fable 5 leads by 2.9 capability points (≈95% interval +0.6 to +5.2): a clear lead over Grok 4.7 · P(higher) ≈ 99%

Wins 90 of 117 shared benchmarks (21 lost, 6 tied)

Capability vs. relative cost
Claude Fable 5Grok 4.7Closed 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 5vsGrok 4.7capability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins90 – 21 · 117 shared, 6 tied

Comparing matching setups only (41 benchmarks excluded) · suites counted once
← Claude Fable 5Grok 4.7 →

Claude Fable 5vsGrok 4.7wins by categoryData Nov 15, 2026Data: Mercor, Arena, Vals AI and 39 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Grok 4.7 minus Claude Fable 5.

FrontierMath Tier 4 (v2) - PrivateMath90.217.1−73.1
ProofBench v1.1Math95.026.0−69.0
KernelBench Mega Normalized (RTX PRO 6000)Code69.418.0−51.4
Medical Long Context Reasoning (MLCR-AA)Health & Medicine64.415.0−49.4
Olam Social Poker EloGames15531508−48.9
Vending-Bench 2Games453010537+36.4
FrontierMath Tiers 1-3 (v2) - PrivateMath87.053.0−34.0
ARC-AGI-2General QA89.261.4−27.8
LLM Creative Story WritingChat & Writing2.90.4−27.4
ArXivMath 05/2026Math87.560.8−26.7
SWE-bench Verified Extended (Mercor held-out)Code78.755.3−23.4
Mystery Game PuzzlesGames52.029.0−23.0
ALE-BenchCode24631823−22.7
Humanity's Last Exam DiamondGeneral QA45.623.6−22.0
TaxCalcBench TY25Finance & Accounting30.08.0−22.0
Toloka Arena (v1)Agents60.939.3−21.6
ChartographyMultimodal34.814.7−20.1
FrontierCUA pass@1Agents40.020.0−20.0
Terminal-Bench 4.0 (Artificial Analysis)Code42.425.8−16.7
SWE-bench VerifiedCode95.579.1−16.4
Extended NYT Connections 940 (Quadratic v1)Games92.776.8−15.9
Vision Arena OverallMultimodal13091249−15.5
SWE Atlas Codebase QnA Extended (Mercor held-out)Code0.015.3+15.3
ArXivMath 06/2026Math85.470.1−15.3
LLM Debate BenchmarkChat & Writing17271620−15.2
AdvancedIFChat & Writing65.280.2+15.0
Furniture Assembly BenchmarkMultimodal35.820.8−15.0
SWE Atlas - Codebase QnACode33.147.0+13.9
Conceptual Reasoning IndexIndexes72.763.1−13.1
LMArena Agent Arena Net ImprovementAgents8.04.0−13.1
ReactBench (39 tasks)Code44.632.3−12.3
LMCAGeneral QA61.149.4−11.8
SAGEMath51.940.8−11.1
ACCoRDGeneral QA85.074.0−10.9
YC-BenchGames19775731666389−10.9
LMArena Text Arena EloChat & Writing15041442−10.8
ComplexConstraintsChat & Writing38.148.9+10.8
SWE-bench MultilingualCode94.884.2−10.6
APEX-AccountingFinance & Accounting56.446.7−9.7
LiveBench 2.0Agents83.674.1−9.5
GPQA DiamondGeneral QA83.392.7+9.4
MMLU ProGeneral QA79.488.8+9.4
GBENCH One-shot Coding GScoreGames65.656.3−9.3
Ruby LLM Benchmarks — Program FixerCode77.268.0−9.3
RuneBenchGames6.05.4−9.2
LMArena Agent Arena Net Improvement (Steering Burden)Agents7.32.6−9.0
Humanity's Last Exam (Text-Only)General QA51.543.1−8.4
ARC-AGI-1 Semi-PrivateGeneral QA98.590.2−8.3
Roboflow Vision Evals (October 2026, five tasks)Multimodal71.763.5−8.2
CadQueryEval (Corrected Geometry)Code80.088.0+8.0
Concept Synth INDUCTION Challenge64 Round 1General QA42.234.4−7.8
Senior SWE-BenchCode34.727.4−7.3
MedXpertQA MM (200-question set)Health & Medicine76.269.0−7.2
RedlineBenchLegal42.449.5+7.1
Concept Synth INDUCTION Challenge100 Round 1General QA31.024.0−7.0
GDP.pdfProfessional Work29.822.8−7.0
Terminal-Bench 4.0Code44.537.6−7.0
Arena Image-to-WebDevCode16211589−6.9
Roboflow Vision Evals (September 2026 protocol)Multimodal78.771.9−6.8
BrowseComp (Mercor 130-question subset)Search82.575.8−6.7
Excel Modeling BenchmarkFinance & Accounting73.767.0−6.7
AA-Briefcase v1.1 EloAgents15361657+6.7
MedCodeHealth & Medicine56.149.6−6.5
KernelBench CUDA Normalized (RTX PRO 6000)Code51.645.2−6.4
AA-AnalystAgent (pass^5)Professional Work48.855.0+6.3
Blueprint-Bench 2Multimodal0.40.3−6.1
Chess Puzzles (task version 2.0)Games47.041.0−6.0
APEX-Agents 1.1Professional Work74.568.7−5.8
FrontierCode 1.1 Extended - Pass RateCode70.965.2−5.6
LMArena Agent Arena Praise vs ComplaintAgents14.33.1−5.6
FrontierCode 1.1 Extended - ScoreCode64.959.4−5.6
DAYJOB: FinanceFinance & Accounting9.014.5+5.5
APEX-SWECode58.953.4−5.5
LMArena Agent Arena SteerabilityAgents11.40.5−5.5
MMMU ProMultimodal72.778.2+5.5
LMArena Agent Arena Steering BurdenAgents7.9-2.3−5.1
BrowseComp Extended (Mercor held-out)Search44.539.5−5.0
Terminal-Bench 2.1 Extended (Mercor held-out)Code16.811.8−5.0
Professional Reasoning Benchmark - LegalLegal52.647.6−4.9
BrokenArXiv 06/2026Math47.852.8+4.9
MedXpertQA MM Extended (Mercor held-out)Health & Medicine56.952.0−4.9
BrokenArXiv 05/2026Math44.540.0−4.5
Physical-Engineering's Last Exam (hidden test)Professional Work34.530.2−4.3
Vibe Code Bench v1.1Code90.486.2−4.2
LegalBenchLegal88.684.4−4.2
Vals RSI Index v1.1 (four tasks)Agents24.320.2−4.1
GeneBench-Pro Public Release SubsetScience14.010.0−4.0
ChessBench CoherenceGames82.686.3+3.7
SWE-TogetherCode68.865.1−3.7
ChessBench AccuracyGames77.174.0−3.3
Terminal-Bench 2.1Code88.085.0−3.0
GameBench 2Games75.272.3−2.9
CharXiv (descriptive and reasoning combined)Multimodal89.192.0+2.9
PLE (Physical-Engineering's Last Exam)Professional Work34.932.0−2.9
AA-LCR Extended (Mercor held-out)Search71.168.2−2.9
DAYJOB: HealthcareHealth & Medicine5.68.4+2.8
CharXiv Extended (Mercor held-out)Multimodal80.778.2−2.5
DTBenchGeneral QA98.496.0−2.4
EnterpriseBench: CoreCraftProfessional Work70.368.2−2.1
LMArena Code Arena WebDevCode16251641+2.0
OTIS Mock AIME 2024-2025Math100.098.1−1.9
ChessBench EloGames10461076+1.8
Terminal-Bench Science 0.1 (Vals)Science12.911.4−1.4
MineBench — Bradley-Terry RatingGames18691848−1.3
LMArena Agent Arena Bash RecoveryAgents9.67.2−1.2
DeepSWE v1.1Code69.971.0+1.1
Tax Agent BenchFinance & Accounting69.869.0−0.8
APEX-1 (Mercor, mean score)Professional Work66.065.4−0.6
LMArena Agent Arena Bash ReliabilityAgents10.49.3−0.6
LMArena Agent Arena Tool HallucinationAgents0.41.0+0.3
LMArena Agent Arena Confirmed SuccessAgents5.15.10.0
AIME 2025Math100.0100.00.0
Bug Hunt BenchCode27.627.60.0
Eval Connections One-Shot with TrapsGames85.085.00.0
FrontierMath ErdősMath0.00.00.0
ObviousBenchGeneral QA99.399.30.0
ProgramBench (Mercor 100-task subset)Code0.00.00.0

Claude Fable 5vsGrok 4.7normalised score marginsData Nov 15, 2026Data: Mercor, Arena, Vals AI and 39 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…