Atlas

Compare Models

GPT-6 Astra vs Muse Spark 1.3

GPT-6 Astra leads by 6.1 capability points (≈95% interval +3.7 to +8.4): a clear lead over Muse Spark 1.3 · P(higher) ≈ 100%

Wins 147 of 233 shared benchmarks (73 lost, 13 tied)

Capability vs. relative cost
GPT-6 AstraMuse 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 AstravsMuse Spark 1.3capability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins147 – 73 · 233 shared, 13 tied

Comparing matching setups only (108 benchmarks excluded) · suites counted once
← GPT-6 AstraMuse Spark 1.3 →

GPT-6 AstravsMuse Spark 1.3wins by categoryData Nov 15, 2026Data: Vals AI, Artificial Analysis, Mercor and 55 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 Astra.

RSI Arena Stage 1 human ratingAgents1095938−75.5
VoxelBench Text RatingGames26251616−71.9
AA-PersonalityBench Informativeness indexChat & Writing29.6100+71.2
Olam Social Poker EloGames15501494−60.9
Harvey's Legal Agent Benchmark - Corporate GovernanceLegal0.060.0+60.0
Harvey's Legal Agent Benchmark - Data Privacy CybersecurityLegal0.060.0+60.0
AA-PersonalityBench Slop indexChat & Writing16.275.3+59.6
CyberGym-E2E-AA (Artificial Analysis)Cybersecurity0.059.5+59.5
Mystery Game PuzzlesGames84.025.0−59.0
Design Creativity BenchCode65.547.2−58.0
BrokenArXiv 05/2026Math91.036.5−54.5
BabaIsBench Stage 0: The IntroGames100.046.0−54.0
AA-PersonalityBench Transparency indexChat & Writing94.341.0−53.8
SRE Bench — Fully SolvedCode56.93.8−53.1
Terminal-Bench Science 0.1 (Vals)Science65.714.3−51.4
FrontierMath Tier 4 (v2) - PrivateMath97.646.3−51.3
Terminal-Bench-Science 0.1Science54.83.8−51.0
BrokenArXiv 04/2026Math97.549.2−48.4
BrokenArXiv 06/2026Math99.153.2−45.8
RiemannBenchMath72.028.0−44.0
IOI (Vals v2)Code100.056.6−43.4
ChartographyMultimodal71.027.6−43.4
Terminal-Bench 4.0 (Artificial Analysis)Code59.616.7−42.9
FrontierSWE V2 Worst@5Code55.112.6−42.5
ALE-BenchCode31641981−41.9
CheatBenchSafety45.086.5−41.5
TextQuests Game ProgressGames74.233.0−41.2
ProofBench v1.1Math99.058.0−41.0
Harvey's Legal Agent Benchmark - Bankruptcy RestructuringLegal0.040.0+40.0
Harvey's Legal Agent Benchmark - Capital MarketsLegal0.040.0+40.0
Harvey's Legal Agent Benchmark - Emerging Companies Venture CapitalLegal0.040.0+40.0
Harvey's Legal Agent Benchmark - Healthcare Life SciencesLegal0.040.0+40.0
Harvey's Legal Agent Benchmark - ImmigrationLegal0.040.0+40.0
TaxCalcBench TY25Finance & Accounting44.06.0−38.0
Chess Puzzles (task version 2.0)Games74.037.0−37.0
WeirdML v3Science42.26.7−35.6
Hedge-BenchFinance & Accounting44.29.1−35.1
NoCoT-Bench NCRI 15.2General QA15998.3−33.4
Code2Video-AA EloCode14271019−33.1
ChessBench EloGames17551201−33.0
HoneyBench v0.2 (hack rate)Safety50.883.6−32.8
AA-PersonalityBench Conviction indexChat & Writing44.877.2+32.7
LLM Creative Story WritingChat & Writing3.40.6−32.6
FrontierSWE V2 Best@5Code73.040.9−32.1
CyberBench v1.1 OverallCybersecurity41.172.7+31.7
Decision-Bench v1.1Professional Work51.519.8−31.6
Humanity's Last Exam DiamondGeneral QA63.833.7−30.1
Terminal-Bench 4.0Code57.127.8−29.3
GBENCH One-shot Coding GScoreGames79.252.7−26.5
MMMU-Pro Extended (Mercor held-out)Multimodal60.033.8−26.2
GameBench 2Games94.368.5−25.8
Last Translation Benchmark (LTBv1-eval, blind)Chat & Writing44.018.4−25.6
DeepsecBench-AA (Artificial Analysis)Cybersecurity36.912.5−24.4
ArXivMath 04/2026Math90.866.7−24.2
LMArena Agent Arena Net Improvement (Steering Burden)Agents14.52.2−23.3
Snorkel Agentic Coding 2.0Code47.625.0−22.6
AutoResearchExam (hidden-test AUARC)Agents0.60.4−22.1
SheetBench-50Finance & Accounting64.742.8−21.9
ArXivMath 05/2026Math95.073.3−21.7
LMArena Agent Arena Net ImprovementAgents10.43.8−21.6
AA-PersonalityBench Warmth indexChat & Writing75.396.2+21.2
AA-PersonalityBench Imagination indexChat & Writing77.298.1+21.2
AutomationBench 1.0.6 (Zapier)Agents41.420.7−20.7
Code MigrationCode67.747.4−20.3
Artificial Analysis Intelligence Index v5.0.1Indexes56.936.6−20.2
Harvey's Legal Agent Benchmark - Antitrust CompetitionLegal0.020.0+20.0
Harvey's Legal Agent Benchmark - Arbitration International Dispute ResolutionLegal0.020.0+20.0
Harvey's Legal Agent Benchmark - Banking FinanceLegal0.020.0+20.0
Harvey's Legal Agent Benchmark - Energy Natural ResourcesLegal0.020.0+20.0
Harvey's Legal Agent Benchmark - Funds Asset ManagementLegal0.020.0+20.0
Harvey's Legal Agent Benchmark - International Trade SanctionsLegal0.020.0+20.0
Harvey's Legal Agent Benchmark - Real EstateLegal0.020.0+20.0
Harvey's Legal Agent Benchmark - Structured Finance SecuritizationLegal0.020.0+20.0
AA-3D v1.0 EloCode11681032−19.9
EnterpriseBench: CoreCraftProfessional Work42.661.9+19.3
FrontierMath Tiers 1-3 (v2) - PrivateMath93.774.4−19.3
FrontierCode 2.0 (Artificial Analysis)Code42.423.1−19.3
Artificial Analysis Intelligence Index v5.0Indexes55.336.4−18.8
sudo L7 (October 2026 rescoring)Code31.112.8−18.3
ArXivMath 06/2026Math94.476.7−17.7
MysteryMechanismScience53.236.0−17.1
Humanity's Sixth SenseMultimodal53.637.4−16.2
BuseyBench SVGMultimodal7.35.7−16.1
CadQueryEval (Corrected Geometry)Code100.084.0−16.0
Legal Research BenchLegal39.455.3+15.9
SnorkelLegalLegal20.936.5+15.6
ITBench-AAAgents48.633.2−15.3
DAYJOB: Data ScienceProfessional Work21.36.0−15.3
CritPt v2 (agentic, avg@5)Science83.168.0−15.1
LMArena Agent Arena Praise vs ComplaintAgents31.71.7−15.0
Finance Agent v2 - General Qualitative AnalysisFinance & Accounting68.483.4+14.9
DAYJOB: LawLegal36.922.8−14.1
HardGeoBench Avg DistanceMultimodal14781022+14.0
APEX-Agents 1.1Professional Work75.161.3−13.8
AA-AnalystAgent (pass^5)Professional Work51.337.5−13.7
Creative Writing v3 EloChat & Writing21771907−13.7
AA-PersonalityBench Expressiveness indexChat & Writing58.171.4+13.5
Professional Reasoning Benchmark - LegalLegal48.461.6+13.2
Extended NYT Connections 940 (Quadratic v1)Games98.185.1−13.0
NoCoT-Bench NCKI (kspine v3)General QA131113−12.7
Professional Reasoning Benchmark - FinanceFinance & Accounting47.559.5+12.0
DAYJOB: AccountingFinance & Accounting30.018.0−12.0
Finance Agent v2 - Financial ModelingFinance & Accounting21.633.5+11.9
DAXBench v2Code96.885.1−11.7
LiveBench 2.0Agents86.174.8−11.4
Artificial Analysis Cyber Index v1Indexes33.444.3+10.9
Arena Alignment IndexSafety87.877.1−10.7
AutomationBench-AA 1.0.6Agents68.557.9−10.6
Conceptual Reasoning IndexIndexes73.966.2−10.6
DistressBenchSafety74.384.9+10.6
LMCAGeneral QA64.453.9−10.5
MedXpertQA MM (200-question set)Health & Medicine86.876.3−10.5
Humanity's Last ExamGeneral QA53.643.2−10.4
Snorkel SWE-bench CLICode12.42.2−10.2
Eval Connections One-Shot with TrapsGames94.084.0−10.0
GeneBench-Pro Public Release SubsetScience36.026.0−10.0
ChessBench AccuracyGames85.076.0−9.6
APEX-SWECode48.639.1−9.5
Finance Agent v2 - ComparablesFinance & Accounting37.847.2+9.4
AA-Omniscience IndexIndexes43.725.0−9.4
Westworld Finance Diligence BenchFinance & Accounting54.545.2−9.3
Finance Agent v2 - All PassFinance & Accounting40.349.5+9.3
Tax Agent BenchFinance & Accounting63.372.4+9.1
Epoch Capabilities Index (ECI)Indexes169158−9.0
Artificial Analysis Progress Index v1Indexes350324−8.8
HardGeoBench <=100miMultimodal46.355.0+8.8
MentalHealthBenchSafety57.348.6−8.7
MA-BenchProfessional Work53.461.8+8.4
Finance Agent v2 - Disclosure AnalysisFinance & Accounting61.770.1+8.4
Medical Long Context Reasoning (MLCR-AA)Health & Medicine35.043.3+8.3
GDPval Extended (Mercor held-out rubric score)Professional Work64.155.9−8.2
Finance Agent v2 - Earnings AnalysisFinance & Accounting69.577.2+7.7
HardGeoBench Country AccuracyMultimodal52.560.0+7.5
𝜏³-BankingAgents43.150.5+7.4
Gray Swan IPI Attack Success (k=15)Safety8.515.9−7.4
Harvey HLAB — Criteria Pass RateLegal87.594.7+7.3
Vibe Code Bench 1–100Code27.620.5−7.2
Roboflow Vision Evals (September 2026 protocol)Multimodal86.679.8−6.8
Design Arena Overall Code CategoriesCode13891348−6.8
Snorkel Digital Work IndexIndexes23.016.3−6.7
DAYJOB: FinanceFinance & Accounting21.514.8−6.7
ACCoRDGeneral QA81.775.1−6.6
DrugDiscoveryBenchScience68.762.2−6.5
FrontierCode 1.1 Main (Artificial Analysis)Code47.040.6−6.3
Legacy-Bench (Fireworks, avg@3)Code55.762.0+6.3
Ramp Accounting BenchFinance & Accounting48.742.5−6.2
LMArena Agent Arena Steering BurdenAgents14.82.6−6.1
ComplexConstraintsChat & Writing57.751.9−5.8
Harvey LAB (Mercor run, undisclosed ~213-task cohort)Legal9.215.0+5.8
ChessBench CoherenceGames97.091.2−5.8
AA-PersonalityBench Organization indexChat & Writing69.563.8−5.8
CritPt (avg@5)Science31.726.0−5.7
GDP.pdf (Artificial Analysis, All-pass)Professional Work32.226.6−5.6
GDPval-AA v2Professional Work15551662+5.3
LMArena Agent Arena Confirmed SuccessAgents14.03.7−5.2
Document ArenaProfessional Work14671473+5.1
Shroombench (mushroom identification, top-1)Multimodal54.049.0−5.0
Finance Agent v2 - PrecedentsFinance & Accounting30.335.2+4.9
GPQA DiamondGeneral QA96.291.3−4.9
Artificial Analysis Intelligence Index v4.3Indexes52.848.2−4.6
APEX-1 (Mercor, mean score)Professional Work69.373.9+4.6
Artificial Analysis Intelligence Index v4.3.2Indexes52.748.1−4.6
Finance Agent v2 - AdjustmentsFinance & Accounting48.352.8+4.5
APEX-AccountingFinance & Accounting57.953.5−4.4
Finance Agent v2 - Market AnalysisFinance & Accounting65.569.9+4.4
Excel Modeling BenchmarkFinance & Accounting71.767.4−4.3
GDPval-AA v2.1Professional Work15871681+4.1
LMArena Text Arena EloChat & Writing14741496+3.9
DAYJOB: HealthcareHealth & Medicine11.67.8−3.8
Terminal-Bench 2.1Code89.185.4−3.7
Financial Coaching BenchFinance & Accounting67.571.0+3.5
CharXiv Extended (Mercor held-out)Multimodal85.181.8−3.3
PCT Political Inconsistency (CAIS)Safety37.440.6−3.2
SciCode 1.0.1 (Artificial Analysis)Code56.559.7+3.2
AA-LCR v1.1Search80.083.0+3.0
ProgramBenchCode5.52.5−3.0
WellnessBenchSafety64.061.0−3.0
LMArena Agent Arena Bash ReliabilityAgents7.82.0−2.9
Immigration BenchLegal67.765.0−2.7
HANDBOOK.md Agents v1.1Professional Work31.433.9+2.5
HardGeoBench <=10miMultimodal36.338.8+2.5
CWE-Bench-AA (Artificial Analysis)Cybersecurity63.360.8−2.5
SpeciEval (Pooled English Runs)Safety96.598.7+2.2
SpeciEval (Refusal-Aware Retries, Historical)Safety96.598.8+2.2
micro1 PharmaBenchHealth & Medicine16.113.9−2.2
CharXiv (descriptive and reasoning combined)Multimodal96.193.9−2.2
Vals Index v2Indexes66.664.5−2.1
Vision Arena OverallMultimodal12831291+2.1
LLM Debate BenchmarkChat & Writing16061593−1.8
OpenAI MRCR v2 8-needle - 512K-1MGeneral QA96.398.1+1.8
SimpleBenchGeneral QA83.681.8−1.8
SWE-TogetherCode60.662.4+1.8
SnorkelRevOpsProfessional Work10.912.6+1.7
SnorkelFinance 2.0Finance & Accounting11.610.0−1.6
SnorkelUnderwrite 2.0Finance & Accounting21.219.6−1.6
AA-Briefcase EloAgents15621589+1.6
HiL-Bench (Human-in-Loop Benchmark)Agents65.366.9+1.6
OpenAI MRCR v2 8-needle - 256K-512KGeneral QA100.098.5−1.5
Creative Writing v3 Rubric ScoreChat & Writing15.915.6−1.5
LMArena Agent Arena SteerabilityAgents-1.41.6+1.5
DeepSWE v1.1Code74.175.4+1.3
DAYJOB: TaxFinance & Accounting4.83.6−1.2
Ruby LLM Benchmarks — Program FixerCode75.174.1−1.0
ForecastBench Preliminary Dataset Brier IndexForecasting62.961.9−1.0
ForecastBench Tournament Overall Brier IndexForecasting66.965.9−1.0
LegalReasonBenchLegal39.540.4+0.9
OTIS Mock AIME 2024-2025Math100.099.2−0.8
DTBenchGeneral QA97.396.5−0.8
Finance Agent v2 - General Quantitative AnalysisFinance & Accounting78.878.0−0.8
MedXpertQA MM Extended (Mercor held-out)Health & Medicine60.860.1−0.7
ObviousBenchGeneral QA100.099.3−0.7
HackerRank-BenchCode43.542.9−0.7
Bench to the Future 3 — September Panel — PooledForecasting0.10.1−0.6
CVDP (517-problem set)Code54.954.3−0.6
SnorkelManufacturingProfessional Work10.010.5+0.5
GDP.xlsxProfessional Work22.922.6−0.3
Harvey LAB-AA v1.1 (hallucination-gated)Legal8.68.9+0.3
LMArena Agent Arena Bash RecoveryAgents5.56.1+0.3
SWE-bench Multimodal (Vals v1, 480 Tasks)Code43.343.1−0.2
LMArena Agent Arena Tool HallucinationAgents1.00.8−0.1
Agents' Last Exam (ALE)Agents32.232.20.0
AIME 2025Math100.0100.00.0
Harvey's Legal Agent Benchmark - Corporate M&ALegal20.020.00.0
Harvey's Legal Agent Benchmark - Employment LaborLegal0.00.00.0
Harvey's Legal Agent Benchmark - Environmental ESGLegal0.00.00.0
Harvey's Legal Agent Benchmark - InsuranceLegal0.00.00.0
Harvey's Legal Agent Benchmark - Intellectual PropertyLegal0.00.00.0
Harvey's Legal Agent Benchmark - Litigation Dispute ResolutionLegal0.00.00.0
Harvey's Legal Agent Benchmark - TaxLegal0.00.00.0
Harvey's Legal Agent Benchmark - Trusts Estates Private ClientLegal20.020.00.0
Harvey's Legal Agent Benchmark - White Collar Defense InvestigationsLegal0.00.00.0
Longform Creative WritingChat & Writing82.882.80.0
Visa Agentic Commerce BenchmarkAgents64.464.40.0

GPT-6 AstravsMuse Spark 1.3normalised score marginsData Nov 15, 2026Data: Vals AI, Artificial Analysis, Mercor and 55 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…