Atlas

Compare Models

Claude Fable 5 vs GPT-6.1 Sol

GPT-6.1 Sol leads by 2.8 capability points (≈95% interval +0.3 to +5.3): a clear lead over Claude Fable 5 · P(higher) ≈ 99%

Wins 81 of 117 shared benchmarks (33 lost, 3 tied)

Capability vs. relative cost
Claude Fable 5GPT-6.1 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 5vsGPT-6.1 Solcapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins33 – 81 · 117 shared, 3 tied

Comparing matching setups only (96 benchmarks excluded) · suites counted once
← Claude Fable 5GPT-6.1 Sol →

Claude Fable 5vsGPT-6.1 Solwins by categoryData Nov 15, 2026Data: Mercor, Center for AI Safety, Arena and 36 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is GPT-6.1 Sol minus Claude Fable 5.

VCT-Refusal (CAIS)Safety100.016.1−83.9
ActiveVisionMultimodal7.171.8+64.7
Concept Synth INDUCTION Challenge100 Round 1General QA31.090.0+59.0
HPCT Refusal (CAIS)Safety100.042.0−58.0
KernelBench Mega Normalized (RTX PRO 6000)Code69.414.1−55.3
Concept Synth INDUCTION Challenge64 Round 1General QA42.295.3+53.1
BrokenArXiv 06/2026Math47.8100.0+52.2
BrokenArXiv 05/2026Math44.590.0+45.5
Furniture Assembly BenchmarkMultimodal35.880.0+44.2
ClockBench AccuracyMultimodal37.274.4+37.2
YC-BenchGames1977573936267−36.4
SWE-bench Verified Extended (Mercor held-out)Code78.745.3−33.4
Clocktower BenchGames15341764+33.0
VoxelBench Text RatingGames21012550+32.0
SWE Atlas Codebase QnA Extended (Mercor held-out)Code0.031.3+31.3
ChessBench EloGames10461562+30.7
Medical Long Context Reasoning (MLCR-AA)Health & Medicine64.433.9−30.6
Terminal-Bench 2.1 Extended (Mercor held-out)Code16.845.8+29.0
Mystery Game PuzzlesGames52.080.0+28.0
MMMU-Pro Extended (Mercor held-out)Multimodal37.661.3+23.7
BrowseComp Extended (Mercor held-out)Search44.567.7+23.2
Chess Puzzles (task version 2.0)Games47.068.0+21.0
Agent Red Teaming k=100 (CAIS)Safety39.418.8+20.6
EnterpriseBench: CoreCraftProfessional Work70.350.3−20.0
TextQuests Average Cumulative HarmSafety23.729.3−19.0
GeneBench-Pro Public Release SubsetScience14.033.0+19.0
ALE-BenchCode24632990+18.6
AdvancedIFChat & Writing65.282.4+17.2
GameBench 2Games75.292.0+16.7
NoCoT-Bench NCRI 15.2General QA119149+16.5
SpatialViz-BenchMultimodal73.990.3+16.3
LisanBench Path LengthGames136808736−15.4
TextQuests Game ProgressGames56.171.4+15.4
SWE Atlas - Codebase QnACode33.148.4+15.3
OpenAI internal computer-use safety - adverse outcome rateSafety18.34.3+14.0
Terminal-Bench 4.0 (Artificial Analysis)Code42.456.1+13.6
GBENCH One-shot Coding GScoreGames65.679.0+13.5
SUPERChemScience64.477.8+13.4
Humanity's Last Exam DiamondGeneral QA45.658.8+13.2
BrowseComp (Mercor 130-question subset)Search82.595.4+12.9
MMMU ProMultimodal72.785.1+12.4
Humanity's Sixth SenseMultimodal34.546.6+12.1
APEX-SWECode58.946.9−12.0
MMLU ProGeneral QA79.491.0+11.6
Olam Social Poker EloGames15531543−10.9
SWE-TogetherCode68.858.3−10.5
ERQAMultimodal67.878.2+10.4
Roboflow Vision Evals (October 2026, five tasks)Multimodal71.781.7+10.0
SWE-bench MultilingualCode94.884.9−9.9
Agents' Last Exam Overall ScoreAgents48.758.6+9.9
FrontierMath Tier 4 (v2) - PrivateMath90.2100.0+9.8
LMArena Agent Arena Net Improvement (Steering Burden)Agents7.312.5+9.7
MedXpertQA MM (200-question set)Health & Medicine76.285.7+9.5
ArXivMath 06/2026Math85.494.4+9.0
DAYJOB: FinanceFinance & Accounting9.018.0+9.0
BullshitBench v2 LeaderboardSafety56.065.0+9.0
Terminal-Bench 4.0Code44.553.4+8.9
HieroglyphBenchMultimodal22.931.4+8.5
KernelBench CUDA Normalized (RTX PRO 6000)Code51.643.2−8.4
SWE-bench VerifiedCode95.587.3−8.2
TaxCalcBench TY25Finance & Accounting30.038.0+8.0
MedXpertQA MM Extended (Mercor held-out)Health & Medicine56.964.7+7.8
Tax Agent BenchFinance & Accounting69.862.3−7.5
ArXivMath 05/2026Math87.595.0+7.5
MedCodeHealth & Medicine56.148.8−7.2
CharXiv (descriptive and reasoning combined)Multimodal89.196.0+6.9
SciCode (main problems)Science21.514.6−6.9
Roboflow Vision Evals (September 2026 protocol)Multimodal78.785.5+6.8
FrontierCUA pass@1Agents40.046.7+6.7
FrontierMath Tiers 1-3 (v2) - PrivateMath87.093.7+6.7
LMArena Agent Arena Praise vs ComplaintAgents14.327.5+6.6
Humanity's Last ExamGeneral QA57.851.3−6.5
ProgramBench (Mercor 100-task subset)Code0.06.3+6.3
MindCube TinyMultimodal83.189.1+6.0
HiL-Bench (Human-in-Loop Benchmark)Agents56.362.3+6.0
LisanBench Difficulty-WeightedGames45624004−6.0
MASKSafety92.898.3+5.5
SAGEMath51.946.5−5.4
IntPhys 2Multimodal67.172.3+5.3
GDP.pdfProfessional Work29.835.0+5.2
ARC-AGI-2General QA89.294.2+5.0
ChessBench CoherenceGames82.687.5+4.9
FrontierMath ErdősMath0.04.4+4.4
CharXiv Extended (Mercor held-out)Multimodal80.785.1+4.4
Vision Arena OverallMultimodal13091293−4.3
Machiavelli Deception (CAIS)Safety80.278.8+4.2
ACCoRDGeneral QA85.080.9−4.1
ProofBench v1.1Math95.099.0+4.0
ChessBench AccuracyGames77.180.7+3.9
LMArena Text Arena EloChat & Writing15041482−3.8
APEX-Accounting 1.1Finance & Accounting61.064.3+3.3
LMArena Agent Arena Confirmed SuccessAgents5.111.7+3.3
Excel Modeling BenchmarkFinance & Accounting73.770.8−2.9
Extended NYT Connections 940 (Quadratic v1)Games92.795.5+2.8
LMArena Agent Arena Steering BurdenAgents7.913.4+2.8
LMArena Agent Arena Bash ReliabilityAgents10.45.0−2.7
Terminal-Bench 2.1Code88.085.4−2.6
LMArena Agent Arena Bash RecoveryAgents9.64.5−2.5
PCT Political Inconsistency (CAIS)Safety39.837.3+2.5
Ruby LLM Benchmarks — Program FixerCode77.275.0−2.2
DAYJOB: HealthcareHealth & Medicine5.67.6+2.0
DTBenchGeneral QA98.496.5−1.9
LiveBench 2.0Agents83.685.4+1.8
Search ArenaSearch12281230+1.8
LMCAGeneral QA61.162.8+1.6
Vibe Code Bench v1.1Code90.488.9−1.4
APEX-Agents 1.1Professional Work74.573.2−1.3
AA-AnalystAgent (pass^5)Professional Work48.850.0+1.3
AA-Briefcase v1.1 EloAgents15361555+1.1
NoCoT-Bench NCKI (kspine v3)General QA128130+1.0
ObviousBenchGeneral QA99.3100.0+0.7
LMArena Agent Arena Tool HallucinationAgents0.41.0+0.3
Prophet Arena Arena Score (Fixed Context)Forecasting84.984.8−0.1
Conceptual Reasoning IndexIndexes72.772.70.0
AIME 2025Math100.0100.00.0
ARC-AGI-1 Semi-PrivateGeneral QA98.598.50.0
OTIS Mock AIME 2024-2025Math100.0100.00.0

Claude Fable 5vsGPT-6.1 Solnormalised score marginsData Nov 15, 2026Data: Mercor, Center for AI Safety, Arena and 36 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…