Atlas

Compare Models

Claude Fable 5.2 vs GPT-5.6 Sol

Claude Fable 5.2 leads by 7.4 capability points (≈95% interval +4.7 to +10.1): a clear lead over GPT-5.6 Sol · P(higher) ≈ 100%

Wins 50 of 55 shared benchmarks (4 lost, 1 tied)

Capability vs. relative cost
Claude Fable 5.2GPT-5.6 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-5.6 Solcapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins50 – 4 · 55 shared, 1 tied

Comparing matching setups only (18 benchmarks excluded) · suites counted once
← Claude Fable 5.2GPT-5.6 Sol →

Claude Fable 5.2vsGPT-5.6 Solwins by categoryData Nov 15, 2026Data: Artificial Analysis, Epoch AI, Arena and 18 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is GPT-5.6 Sol minus Claude Fable 5.2.

ARC-AGI-3 (Semi-Private)General QA81.27.0−74.2
Humanity's Last Exam DiamondGeneral QA80.230.4−49.8
LMArena Agent Arena Net Improvement (Steering Burden)Agents30.44.6−48.6
EBR-bench v4 (card ban, single agent)Games90.544.8−45.7
FrontierSWE V2Code69.932.2−37.7
LMArena Agent Arena Praise vs ComplaintAgents73.46.3−33.6
Furniture Assembly BenchmarkMultimodal90.056.7−33.3
Mystery Game PuzzlesGames90.058.0−32.0
ChartographyMultimodal73.445.0−28.4
SWE-Bench Pro (Public Dataset)Code93.064.6−28.4
Humanity's Last ExamGeneral QA73.245.5−27.7
HardGeoBench Avg DistanceMultimodal5521429−27.1
Clocktower BenchGames17591583−25.3
HardGeoBench Country AccuracyMultimodal70.047.5−22.5
BioSecBench-Structure V1 (58 evaluations)Science41.420.7−20.7
Chess Puzzles (task version 2.0)Games79.059.0−20.0
GameBench 2Games99.180.1−19.1
HardGeoBench <=100miMultimodal60.041.3−18.7
HardGeoBench <=10miMultimodal48.830.0−18.7
MysteryMechanismScience51.833.3−18.5
CrossSearch (pass^3)Search56.038.0−18.0
FrontierMath Tier 4 (v2) - PrivateMath100.082.9−17.1
ProofBench v1.1Math100.083.0−17.0
AA-Briefcase v1.1 EloAgents17851479−16.8
HieroglyphBenchMultimodal43.627.6−16.0
AutomationBench 1.0.6 (Zapier)Agents44.328.8−15.5
Conceptual Reasoning IndexIndexes79.968.7−15.3
AutomationBench-AA 1.0.6Agents74.260.1−14.1
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine43.930.0−13.9
LMArena Agent Arena Steering BurdenAgents33.05.5−13.7
ACCoRDGeneral QA89.475.9−13.5
Vibe Code Bench v1.1Code93.880.5−13.3
BioMysteryBench (Vals Terminus 2)Science83.771.1−12.6
LMCAGeneral QA70.658.5−12.1
BuseyBench SVGMultimodal7.26.1−10.6
SciCode 1.0.1 (Artificial Analysis)Code67.757.8−10.0
LMArena Agent Arena Confirmed SuccessAgents22.02.3−9.8
GDPval-AA v2.1Professional Work18311619−9.4
CritPt v2 (agentic, avg@5)Science86.377.1−9.1
Epoch Capabilities Index (ECI)Indexes174163−9.1
HealthBench Professional (Length-Adjusted)Health & Medicine68.160.5−7.6
CritPt (avg@5)Science35.728.6−7.1
FrontierMath Tiers 1-3 (v2) - PrivateMath95.189.1−6.0
HealthBenchHealth & Medicine60.155.3−4.8
GPQA DiamondGeneral QA93.989.9−4.0
DTBenchGeneral QA99.295.7−3.5
MacroscopeBenchCode76.077.8+1.8
DeepSWE v1.1Code74.372.7−1.6
LMArena Agent Arena Bash ReliabilityAgents14.411.4−1.5
ARC-AGI-2General QA93.892.5−1.2
ARC-AGI-1 Semi-PrivateGeneral QA98.597.5−1.0
Tax Agent BenchFinance & Accounting67.068.0+0.9
AA-LCR v1.1Search83.384.0+0.7
OTIS Mock AIME 2024-2025Math99.7100.0+0.3
LMArena Agent Arena Tool HallucinationAgents1.01.00.0

Claude Fable 5.2vsGPT-5.6 Solnormalised score marginsData Nov 15, 2026Data: Artificial Analysis, Epoch AI, Arena and 18 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…