Atlas

Compare Models

Claude Fable 5.2 vs GPT-6 Sol

Claude Fable 5.2 leads by 6.5 capability points (≈95% interval +3.7 to +9.2): a clear lead over GPT-6 Sol · P(higher) ≈ 100%

Wins 68 of 71 shared benchmarks (3 lost)

Capability vs. relative cost
Claude Fable 5.2GPT-6 SolClosed 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 5.2vsGPT-6 Solcapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins68 – 3 · 71 shared

Comparing matching setups only (13 benchmarks excluded) · suites counted once
← Claude Fable 5.2GPT-6 Sol →

Claude Fable 5.2vsGPT-6 Solwins by categoryData Nov 15, 2026Data: Artificial Analysis, Epoch AI, Arena and 16 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is GPT-6 Sol minus Claude Fable 5.2.

ARC-AGI-3 (Semi-Private)General QA97.39.5−87.8
AA-PersonalityBench Expressiveness indexChat & Writing81.033.4−48.1
LMArena Agent Arena Net Improvement (Steering Burden)Agents30.46.9−44.3
AA-PersonalityBench Organization indexChat & Writing54.312.4−42.3
AA-PersonalityBench Transparency indexChat & Writing81.039.1−42.3
AA-PersonalityBench Imagination indexChat & Writing67.629.6−38.5
EBR-bench v4 (card ban, single agent)Games90.553.3−37.1
AA-PersonalityBench Informativeness indexChat & Writing46.710.5−36.5
Humanity's Last Exam DiamondGeneral QA80.245.2−35.0
Mystery Game PuzzlesGames90.056.0−34.0
SimpleQA Verified (Epoch anti-abstention)General QA94.060.7−33.3
LMArena Agent Arena Praise vs ComplaintAgents73.410.0−31.7
Furniture Assembly BenchmarkMultimodal90.058.3−31.7
HardGeoBench Avg DistanceMultimodal5521497−29.2
AA-PersonalityBench Warmth indexChat & Writing82.954.3−28.8
Humanity's Last Exam (Text-Only)General QA74.346.3−28.0
BioSecBench-Structure V1 (58 evaluations)Science41.413.8−27.6
AA-PersonalityBench Conviction indexChat & Writing75.350.5−25.0
HardGeoBench Country AccuracyMultimodal70.046.3−23.7
Chess Puzzles (task version 2.0)Games79.056.0−23.0
HardGeoBench <=100miMultimodal60.037.5−22.5
HardGeoBench <=10miMultimodal48.826.3−22.5
MysteryMechanismScience51.830.2−21.6
AA-3D v1.0 EloCode12711131−20.3
ChartographyMultimodal73.453.6−19.8
GameBench 2Games99.179.4−19.7
APEX-Agents 1.1Professional Work87.668.7−18.9
HieroglyphBenchMultimodal43.626.0−17.6
ProofBench v1.1Math100.083.0−17.0
AA-Briefcase v1.1 EloAgents17851476−17.0
SWE-bench MultilingualCode96.681.7−14.9
GDPval-AA v2.1Professional Work18311514−14.0
Tax Agent BenchFinance & Accounting67.053.0−14.0
FrontierCode 2.0 (Artificial Analysis)Code48.835.6−13.2
τ³-Banking (Mercor run, bm25_grep retrieval)Agents55.743.0−12.7
AutomationBench-AA 1.0.6Agents74.261.7−12.5
Conceptual Reasoning IndexIndexes79.971.0−12.2
HealthBenchHealth & Medicine60.148.2−11.9
LMCAGeneral QA70.659.1−11.5
LMArena Agent Arena Steering BurdenAgents33.010.1−11.4
AutomationBench 1.0.6 (Zapier)Agents44.333.2−11.1
APEX-Accounting 1.1Finance & Accounting70.459.7−10.7
SpireClimbGames19.69.2−10.5
HealthBench ProfessionalHealth & Medicine71.461.2−10.2
SciCode 1.0.1 (Artificial Analysis)Code67.757.6−10.1
FrontierMath Tier 4 (v2) - PrivateMath100.090.0−10.0
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine43.933.9−10.0
BioMysteryBench (Vals Terminus 2)Science83.774.8−8.9
LMArena Agent Arena Confirmed SuccessAgents22.04.5−8.7
Epoch Capabilities Index (ECI)Indexes174164−8.3
DeepSWE v1.1Code74.366.6−7.7
ChessBenchGames69.362.0−7.3
ACCoRDGeneral QA89.482.4−7.0
HealthBench (length-adjusted, Artificial Analysis)Health & Medicine60.454.2−6.2
Vibe Code Bench v1.1Code93.887.8−5.9
Harvey LAB (Mercor run, undisclosed ~213-task cohort)Legal10.04.1−5.9
AA-PersonalityBench Slop indexChat & Writing6.71.0−5.8
CritPt v2 (agentic, avg@5)Science86.380.9−5.4
FrontierMath Tiers 1-3 (v2) - PrivateMath95.189.8−5.3
DualEntry Accounting AI BenchmarkFinance & Accounting81.276.2−5.0
CritPt (avg@5)Science35.730.9−4.9
ARC-AGI-2General QA93.889.6−4.2
Harvey LAB-AA v1.1 (hallucination-gated)Legal7.23.6−3.6
BuseyBench SVGMultimodal7.27.5+3.6
ARC-AGI-1 Semi-PrivateGeneral QA98.595.5−3.0
DTBenchGeneral QA99.297.3−1.9
LMArena Agent Arena Bash ReliabilityAgents14.410.9−1.7
MacroscopeBenchCode78.577.2−1.3
AA-LCR v1.1Search83.383.7+0.3
OTIS Mock AIME 2024-2025Math99.7100.0+0.3
LMArena Agent Arena Tool HallucinationAgents1.01.00.0

Claude Fable 5.2vsGPT-6 Solnormalised score marginsData Nov 15, 2026Data: Artificial Analysis, Epoch AI, Arena and 16 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…