Atlas

Compare Models

GPT-6 Sol vs Muse Spark 1.3

GPT-6 Sol leads by 2.6 capability points (≈95% interval +0.4 to +4.8): a clear lead over Muse Spark 1.3 · P(higher) ≈ 99%

Wins 91 of 147 shared benchmarks (51 lost, 5 tied)

Capability vs. relative cost
GPT-6 SolMuse Spark 1.3Closed 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.

GPT-6 SolvsMuse Spark 1.3capability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins91 – 51 · 147 shared, 5 tied

Comparing matching setups only (70 benchmarks excluded) · suites counted once
← GPT-6 SolMuse Spark 1.3 →

GPT-6 SolvsMuse Spark 1.3wins by categoryData Nov 15, 2026Data: Artificial Analysis, Vals AI, Mercor and 38 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Muse Spark 1.3 minus GPT-6 Sol.

AA-PersonalityBench Informativeness indexChat & Writing10.5100+90.4
AA-PersonalityBench Slop indexChat & Writing1.075.3+75.0
AA-PersonalityBench Imagination indexChat & Writing29.698.1+69.2
CyberGym-E2E-AA (Artificial Analysis)Cybersecurity0.059.5+59.5
BrokenArXiv 08/2026 (Native Harnesses)Math78.620.2−58.3
VoxelBench Text RatingGames23941616−55.4
Olam Social Poker EloGames15421494−52.2
AA-PersonalityBench Organization indexChat & Writing12.463.8+51.9
BrokenArXiv 05/2026Math87.036.5−50.5
ArXivMath 08/2026 (Native Harnesses)Math93.045.6−47.4
FrontierMath Tier 4 (v2) - PrivateMath90.046.3−43.7
AA-PersonalityBench Warmth indexChat & Writing54.396.2+42.3
BrokenArXiv 06/2026Math95.453.2−42.1
AA-PersonalityBench Expressiveness indexChat & Writing33.471.4+38.5
RiemannBenchMath64.828.0−36.8
DeepsecBench-AA (Artificial Analysis)Cybersecurity46.312.5−33.8
Mystery Game PuzzlesGames56.025.0−31.0
Medical Long Context Reasoning (MLCR-AA)Health & Medicine16.143.3+27.2
AA-PersonalityBench Conviction indexChat & Writing50.577.2+26.9
SRE Bench — Fully SolvedCode30.53.8−26.7
Legal Research BenchLegal28.855.3+26.4
IOI (Vals v2)Code82.656.6−26.1
ChartographyMultimodal53.627.6−26.0
ProofBench v1.1Math83.058.0−25.0
BizAgentBench (strict success)Agents61.338.0−23.3
ALE-BenchCode26371981−23.2
Professional Reasoning Benchmark - LegalLegal38.661.6+23.0
Harvey's Legal Agent BenchmarkLegal1.723.8+22.1
TaxCalcBench TY25Finance & Accounting28.06.0−22.0
Terminal-Bench 4.0Code44.424.7−19.7
Tax Agent BenchFinance & Accounting53.072.4+19.4
Terminal-Bench-Science 0.1Science30.011.0−19.0
Chess Puzzles (task version 2.0)Games56.037.0−19.0
EnterpriseBench: CoreCraftProfessional Work43.161.9+18.8
BuseyBench SVGMultimodal7.55.7−18.2
MMMU-Pro Extended (Mercor held-out)Multimodal51.933.8−18.1
GBENCH One-shot Coding GScoreGames70.152.7−17.4
Terminal-Bench Science 0.1 (Vals)Science31.414.3−17.1
DistressBenchSafety68.084.9+16.9
ArXivMath 05/2026Math90.073.3−16.7
Decision-Bench v1.1Professional Work36.519.8−16.7
LMArena Agent Arena Net ImprovementAgents8.83.8−16.4
ITBench-AAAgents49.433.2−16.2
CadQueryEval (Corrected Geometry)Code100.084.0−16.0
FrontierMath Tiers 1-3 (v2) - PrivateMath89.874.4−15.4
SWE-bench Verified Extended (Mercor held-out)Code39.354.0+14.7
AA-3D v1.0 EloCode11311032−14.5
ArXivMath 06/2026Math91.076.7−14.2
WeirdML v3Science19.76.7−13.1
CritPt v2 (agentic, avg@5)Science80.968.0−12.9
AutomationBench 1.0.6 (Zapier)Agents33.220.7−12.5
FrontierCode 2.0 (Artificial Analysis)Code35.623.1−12.5
dfbench Detection RecallCybersecurity59.347.4−11.9
KernelBench CUDA Normalized (RTX PRO 6000)Code42.030.3−11.7
Humanity's Last Exam DiamondGeneral QA45.233.7−11.5
Creative Writing v3 EloChat & Writing21261907−11.1
GameBench 2Games79.468.5−10.9
Finance Agent v2Finance & Accounting49.060.0+10.9
Harvey LAB (Mercor run, undisclosed ~213-task cohort)Legal4.115.0+10.9
Terminal-Bench 4.0 (Artificial Analysis)Code43.933.3−10.6
Arena Alignment IndexSafety87.677.1−10.5
Code MigrationCode57.247.4−9.8
APEX-1 (Mercor, mean score)Professional Work63.673.0+9.4
τ³-Banking (Mercor run, bm25_grep retrieval)Agents43.052.2+9.2
NoCoT-Bench NCRI 15.2General QA11598.3−9.0
LMArena Agent Arena Net Improvement (Steering Burden)Agents6.92.2−8.9
SimpleBenchGeneral QA73.181.8+8.7
LegalReasonBenchLegal31.940.4+8.5
Financial Coaching BenchFinance & Accounting62.671.0+8.4
MedXpertQA MM Extended (Mercor held-out)Health & Medicine68.060.1−7.9
Artificial Analysis Intelligence Index v5.0.1Indexes44.236.6−7.5
Artificial Analysis Cyber Index v1Indexes36.844.3+7.5
APEX-Agents 1.1Professional Work68.761.3−7.4
GDPval-AA v2.1Professional Work15141681+7.4
ACCoRDGeneral QA82.475.1−7.3
LiveBench 2.0Agents81.974.8−7.1
LMArena Text Arena EloChat & Writing14571496+6.9
Conceptual Reasoning IndexIndexes71.066.2−6.7
LMArena Agent Arena SteerabilityAgents14.51.6−6.5
Humanity's Sixth SenseMultimodal31.237.4+6.2
ChessBench EloGames13031201−6.1
APEX-SWECode45.039.1−5.9
MysteryMechanismScience30.236.0+5.9
Gray Swan IPI Attack Success (k=15)Safety10.115.9−5.8
AA-Briefcase v1.1 EloAgents14761581+5.8
DAYJOB: FinanceFinance & Accounting9.314.8+5.5
GDP.xlsxProfessional Work17.122.6+5.5
MentalHealthBenchSafety53.948.6−5.3
Harvey LAB-AA v1.1 (hallucination-gated)Legal3.68.9+5.3
CyberBench v1.1 OverallCybersecurity78.072.7−5.2
LMCAGeneral QA59.153.9−5.2
APEX-Accounting 1.1Finance & Accounting59.764.9+5.2
Eval Connections One-Shot with TrapsGames89.084.0−5.0
Extended NYT Connections 940 (Quadratic v1)Games90.185.1−5.0
MedXpertQA MM (200-question set)Health & Medicine81.376.3−5.0
WellnessBenchSafety56.061.0+5.0
Ruby LLM Benchmarks — Program FixerCode79.074.1−5.0
Epoch Capabilities Index (ECI)Indexes164158−4.9
CritPt (avg@5)Science30.926.0−4.9
LMArena Agent Arena Bash ReliabilityAgents10.92.0−4.5
DAYJOB: HealthcareHealth & Medicine3.67.8+4.2
NoCoT-Bench NCKI (kspine v3)General QA119113−4.1
LMArena Agent Arena Praise vs ComplaintAgents10.01.7−4.1
Terminal-Bench 2.1Code83.179.0−4.1
Excel Modeling BenchmarkFinance & Accounting71.567.4−4.1
AutomationBench-AA 1.0.6Agents61.757.9−3.8
LMArena Agent Arena Steering BurdenAgents10.12.6−3.8
KernelBench Mega Normalized (RTX PRO 6000)Code2.76.2+3.5
Artificial Analysis Progress Index v1Indexes334324−3.4
CWE-Bench-AA (Artificial Analysis)Cybersecurity64.260.8−3.3
AA-LCR Extended (Mercor held-out)Search70.266.9−3.3
ChessBench CoherenceGames88.291.2+3.0
GDPval Extended (Mercor held-out rubric score)Professional Work58.755.9−2.8
AURORA Multilingual Terminal-BenchCode65.362.5−2.8
Roboflow Vision Evals (September 2026 protocol)Multimodal82.379.8−2.5
Humanity's Last Exam (Text-Only)General QA46.348.7+2.4
SciCode 1.0.1 (Artificial Analysis)Code57.659.7+2.1
Creative Writing v3 Rubric ScoreChat & Writing16.115.6−2.0
CharXiv Extended (Mercor held-out)Multimodal83.881.8−2.0
Vibe Code Bench v1.1Code87.885.9−2.0
Vals Index v2Indexes62.664.5+2.0
AA-PersonalityBench Transparency indexChat & Writing39.141.0+1.9
SWE-TogetherCode60.662.4+1.8
CVDP (517-problem set)Code52.754.3+1.6
ChessBench AccuracyGames77.476.0−1.5
GDP.pdf (Artificial Analysis, All-pass)Professional Work25.226.6+1.4
DeepSWE v1.1Code66.667.8+1.2
ComplexConstraintsChat & Writing50.851.9+1.1
CharXiv (descriptive and reasoning combined)Multimodal95.093.9−1.1
AA-Omniscience IndexIndexes27.125.0−1.1
OTIS Mock AIME 2024-2025Math100.099.2−0.8
DTBenchGeneral QA97.396.5−0.8
ForecastBench Preliminary Dataset Brier IndexForecasting62.661.9−0.7
AA-LCR v1.1Search83.783.0−0.7
Vals Index v2.1Indexes57.558.2+0.6
ProgramBenchCode2.02.5+0.5
LMArena Agent Arena Bash RecoveryAgents7.06.1−0.5
Artificial Analysis Intelligence Index v4.3.2Indexes47.648.1+0.5
LMArena Agent Arena Confirmed SuccessAgents4.53.7−0.4
SpeciEval (Refusal-Aware Retries, Historical)Safety98.598.8+0.3
SpeciEval (Pooled English Runs)Safety98.598.7+0.3
LMArena Agent Arena Tool HallucinationAgents1.00.8−0.1
Agents' Last Exam (ALE)Agents32.232.20.0
AIME 2025Math100.0100.00.0
GeneBench-Pro Public Release SubsetScience26.026.00.0
Longform Creative WritingChat & Writing82.882.80.0
ObviousBenchGeneral QA99.399.30.0

GPT-6 SolvsMuse Spark 1.3normalised score marginsData Nov 15, 2026Data: Artificial Analysis, Vals AI, Mercor and 38 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…