Atlas

Compare Models

Claude Fable 5.2 vs Grok 4.7

Claude Fable 5.2 leads by 9.6 capability points (≈95% interval +7.0 to +12.3): a clear lead over Grok 4.7 · P(higher) ≈ 100%

Wins 77 of 84 shared benchmarks (7 lost)

Capability vs. relative cost
Claude Fable 5.2Grok 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 5.2vsGrok 4.7capability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins77 – 7 · 84 shared

Comparing matching setups only (1 benchmark excluded) · suites counted once
← Claude Fable 5.2Grok 4.7 →

Claude Fable 5.2vsGrok 4.7wins by categoryData Nov 15, 2026Data: Artificial Analysis, Vals AI, Epoch AI and 20 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 5.2.

ARC-AGI-3 (Semi-Private)General QA96.88.4−88.4
FrontierMath Tier 4 (v2) - PrivateMath100.017.1−82.9
ProofBench v1.1Math100.026.0−74.0
HoneyBench v0.2 (hack rate)Safety11.985.1−73.2
Furniture Assembly BenchmarkMultimodal90.020.8−69.2
Mystery Game PuzzlesGames90.029.0−61.0
ChartographyMultimodal73.414.7−58.7
Humanity's Last Exam DiamondGeneral QA80.223.6−56.6
BrokenArXiv 09/2026 (Native Harnesses)Math80.527.7−52.8
LMArena Agent Arena Net Improvement (Steering Burden)Agents30.42.6−52.5
ArXivMath 09/2026 (Native Harnesses)Math96.245.3−50.9
HardGeoBench Avg DistanceMultimodal5522104−47.9
OpenWorldEngineer (Mean@8)Code54.67.0−47.5
FrontierMath Tiers 1-3 (v2) - PrivateMath95.153.0−42.1
DeepSWE v1.1Code74.333.3−41.0
AA-3D v1.0 EloCode1271992−40.5
FrontierSWE V2Code69.929.5−40.4
AA-PersonalityBench Warmth indexChat & Writing82.944.8−38.5
Chess Puzzles (task version 2.0)Games79.041.0−38.0
SimpleQA Verified (Epoch anti-abstention)General QA94.056.0−38.0
Agent Initiative BenchmarkAgents8.24.5−36.6
LMArena Agent Arena Praise vs ComplaintAgents73.43.1−35.2
AA-PersonalityBench Imagination indexChat & Writing67.635.3−32.7
HardGeoBench <=100miMultimodal60.027.5−32.5
ARC-AGI-2General QA93.861.4−32.4
HardGeoBench <=10miMultimodal48.817.5−31.2
Humanity's Last Exam (Text-Only)General QA74.343.1−31.1
CritPt v2 (agentic, avg@5)Science83.755.7−28.0
Bug Hunt BenchCode55.227.6−27.6
GameBench 2Games99.172.3−26.8
MysteryMechanismScience51.825.2−26.6
HardGeoBench Country AccuracyMultimodal70.043.8−26.2
BuseyBench SVGMultimodal7.24.6−26.0
LMArena Code Arena WebDevCode18361641−25.5
AA-PersonalityBench Organization indexChat & Writing54.329.6−25.0
τ³-Banking (Mercor run, bm25_grep retrieval)Agents55.732.6−23.1
AA-PersonalityBench Transparency indexChat & Writing81.058.1−23.1
Conceptual Reasoning IndexIndexes79.963.1−23.1
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine43.921.1−22.8
FrontierCode 2.0 - Pass RateCode54.732.3−22.4
FrontierCode 2.0 - ScoreCode50.228.9−21.3
LMCAGeneral QA70.649.4−21.2
FrontierCode 1.1 Main (Artificial Analysis)Code60.639.6−21.0
SWE-bench Multimodal (Vals v1, 480 Tasks)Code62.342.9−19.4
APEX-Agents 1.1Professional Work87.668.7−18.9
CritPt (avg@5)Science35.718.0−17.7
Vals RSI Index v1.1 (four tasks)Agents37.920.2−17.7
LMArena Agent Arena Steering BurdenAgents33.0-2.3−17.6
ACCoRDGeneral QA89.474.0−15.3
APEX-Accounting 1.1Finance & Accounting70.455.3−15.1
Epoch Capabilities Index (ECI)Indexes174156−14.7
BioMysteryBench (Vals Terminus 2)Science83.769.3−14.4
AA-PersonalityBench Informativeness indexChat & Writing46.733.4−13.5
HLAB-Vals-1.1Legal5.018.3+13.3
GDPxProfessional Work16.43.9−12.6
SWE-bench MultilingualCode96.684.2−12.4
SciCode 1.0.1 (Artificial Analysis)Code67.757.8−10.0
Harvey LAB (Mercor run, undisclosed ~213-task cohort)Legal10.019.5+9.5
Finance Agent v2.1Finance & Accounting69.560.6−8.9
AutomationBench-AA 1.0.6Agents74.265.6−8.6
LMArena Agent Arena Confirmed SuccessAgents22.05.1−8.5
ARC-AGI-1 Semi-PrivateGeneral QA98.590.2−8.3
Ramp Accounting BenchFinance & Accounting52.244.1−8.1
AA-PersonalityBench Conviction indexChat & Writing75.382.9+7.7
Vibe Code Bench v1.1Code93.886.2−7.6
DualEntry Accounting AI BenchmarkFinance & Accounting81.288.1+6.9
Immigration BenchLegal74.068.0−5.9
HealthBenchHealth & Medicine60.154.2−5.9
Excel Modeling Benchmark v1.1Finance & Accounting77.471.5−5.9
AA-PersonalityBench Expressiveness indexChat & Writing81.075.3−5.8
AA-Briefcase v1.1 EloAgents17371636−5.6
MineBench — Bradley-Terry RatingGames19321848−5.1
Vals Finance IndexIndexes63.658.8−4.8
HealthBench (length-adjusted, Artificial Analysis)Health & Medicine60.456.0−4.4
DAXBench v2Code88.392.6+4.3
AA-LCR v1.1Search83.380.0−3.3
DTBenchGeneral QA99.296.0−3.2
LMArena Agent Arena Bash ReliabilityAgents14.49.3−2.6
Tax Agent BenchFinance & Accounting67.069.0+2.0
AA-PersonalityBench Slop indexChat & Writing6.78.6+1.9
OTIS Mock AIME 2024-2025Math99.798.1−1.7
GDPval-AA v2.1Professional Work17511717−1.5
GPQA DiamondGeneral QA93.992.7−1.1
LMArena Agent Arena Tool HallucinationAgents1.01.00.0

Claude Fable 5.2vsGrok 4.7normalised score marginsData Nov 15, 2026Data: Artificial Analysis, Vals AI, Epoch AI and 20 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…