Atlas

Compare Models

Opus 4.8 vs Muse Spark 1.3

Muse Spark 1.3 leads by 1.4 capability points (≈95% interval −0.5 to +3.4): statistically tied · P(higher) ≈ 92%

Wins 82 of 128 shared benchmarks (45 lost, 1 tied)

Capability vs. relative cost
Opus 4.8Muse 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.

Opus 4.8vsMuse Spark 1.3capability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins45 – 82 · 128 shared, 1 tied

Comparing matching setups only (66 benchmarks excluded) · suites counted once
← Opus 4.8Muse Spark 1.3 →

Opus 4.8vsMuse Spark 1.3wins by categoryData Nov 15, 2026Data: Vals AI, Artificial Analysis, Mercor and 39 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 Opus 4.8.

KernelBench Mega Normalized (H100)Code63.812.4−51.4
BabaIsBench Stage 0: The IntroGames96.046.0−50.0
HiL-Bench (Human-in-Loop Benchmark)Agents35.366.9+31.6
Olam Social Poker EloGames15181494−26.1
AutomationBench (Public Set)Agents27.249.6+22.4
DeepSWE v1.1Code54.475.4+21.0
WordleBenchGames94.073.0−21.0
RiemannBenchMath47.228.0−19.2
SimpleBenchGeneral QA64.881.8+17.0
Finance Agent v2 - General Qualitative AnalysisFinance & Accounting66.483.4+17.0
JobBench MainProfessional Work48.464.9+16.5
Bug Hunt BenchCode14.330.7+16.4
ComplexConstraintsChat & Writing35.651.9+16.3
𝜏³-BankingAgents34.250.5+16.3
CadQueryEval (Corrected Geometry)Code68.084.0+16.0
AA-Briefcase EloAgents13161589+15.9
DrugDiscoveryBenchScience46.862.2+15.4
Chess Puzzles (task version 2.0)Games52.037.0−15.0
BrokenArXiv 04/2026Math34.449.2+14.8
SpeciEval (Pooled English Runs)Safety84.398.7+14.4
SpeciEval (Refusal-Aware Retries, Historical)Safety84.398.8+14.4
ChessBench EloGames9611201+14.3
Harvey's Legal Agent BenchmarkLegal9.623.8+14.2
TaxCalcBench TY25Finance & Accounting20.06.0−14.0
Humanity's Sixth SenseMultimodal23.537.4+13.9
BrokenArXiv 05/2026Math23.036.5+13.5
DistressBenchSafety71.484.9+13.4
PCT Political Inconsistency (CAIS)Safety53.840.6+13.1
LMArena Code Arena WebDevCode15551655+13.0
AutomationBench-AA 1.0.6Agents45.657.9+12.3
Terminal-Bench 4.0 (Artificial Analysis)Code21.733.3+11.6
BrokenArXiv 06/2026Math41.753.2+11.6
Legal Research BenchLegal43.855.3+11.5
ChessBench CoherenceGames79.791.2+11.5
Roboflow Vision Evals (September 2026 protocol)Multimodal68.779.8+11.1
Mystery Game PuzzlesGames36.025.0−11.0
Humanity's Last Exam DiamondGeneral QA23.133.7+10.6
APEX-1 (Mercor, mean score)Professional Work62.773.0+10.3
Finance Agent v2 - Financial ModelingFinance & Accounting23.333.5+10.2
LMArena Agent Arena Net ImprovementAgents6.93.8−10.0
FrontierMath Tier 4 (v2) - PrivateMath56.146.3−9.8
GDPval-AA v2.1Professional Work14611681+9.7
EnterpriseBench: CoreCraftProfessional Work52.361.9+9.6
Finance Agent v2 - PrecedentsFinance & Accounting25.635.2+9.6
SWE-bench Verified Extended (Mercor held-out)Code63.354.0−9.3
MedXpertQA MM Extended (Mercor held-out)Health & Medicine51.660.1+8.5
ArXivMath 05/2026Math65.073.3+8.3
GeneBench-Pro Public Release SubsetScience18.026.0+8.0
Gray Swan IPI Attack Success (k=15)Safety8.015.9−7.9
Tax Agent BenchFinance & Accounting64.972.4+7.5
AA-AnalystAgent (pass^5)Professional Work45.037.5−7.5
TextQuests Game ProgressGames40.133.0−7.1
Terminal-Bench 2.1Code71.979.0+7.1
ArXivMath 06/2026Math70.076.7+6.8
Humanity's Last ExamGeneral QA49.843.2−6.6
GDPval-AA v2Professional Work15291662+6.6
HANDBOOK.mdProfessional Work21.915.4−6.5
Finance Agent v2 - General Quantitative AnalysisFinance & Accounting71.678.0+6.4
Artificial Analysis Intelligence Index v4.3.2Indexes41.848.1+6.3
Artificial Analysis Intelligence Index v4.3Indexes42.048.2+6.2
GameBench 2Games62.568.5+6.1
Extended NYT Connections 940 (Quadratic v1)Games91.185.1−6.0
Finance Agent v2 - All PassFinance & Accounting43.549.5+6.0
FrontierMath Tiers 1-3 (v2) - PrivateMath80.074.4−5.6
APEX-AccountingFinance & Accounting48.053.5+5.5
LLM Debate BenchmarkChat & Writing16311593−5.4
KernelBench CUDA Normalized (RTX PRO 6000)Code24.930.3+5.4
SciCode 1.0.1 (Artificial Analysis)Code54.459.7+5.3
CharXiv Extended (Mercor held-out)Multimodal76.781.8+5.1
Finance Agent v2 - Earnings AnalysisFinance & Accounting72.177.2+5.1
BuseyBench SVGMultimodal5.25.7+5.0
APEX-SWECode43.939.1−4.8
Vals RSI Index v1.1 (historical five tasks)Agents23.218.7−4.6
Finance Agent v2 - AdjustmentsFinance & Accounting48.452.8+4.4
Finance Agent v2 - Disclosure AnalysisFinance & Accounting65.770.1+4.4
Vals RSI Index v1.1 (four tasks)Agents24.019.6−4.3
Terminal-Bench Science 0.1 (Vals)Science10.014.3+4.3
ALE-BenchCode21021981−4.3
ArXivMath 04/2026Math62.566.7+4.2
LMArena Agent Arena SteerabilityAgents9.61.6−4.0
ChessBench AccuracyGames72.476.0+3.9
AutomationBench 1.0.6 (Zapier)Agents16.920.7+3.8
MedXpertQA MM (200-question set)Health & Medicine72.576.3+3.8
AA-LCR Extended (Mercor held-out)Search63.266.9+3.7
LMCAGeneral QA57.553.9−3.6
Vals Index v2Indexes60.964.5+3.6
GDP.pdfProfessional Work24.027.6+3.6
Vibe Code Bench v1.1Code82.785.9+3.1
APEX-Agents 1.1Professional Work64.461.3−3.1
NoCoT-Bench NCKI (kspine v3)General QA117113−3.1
Vals Index v2.1Indexes55.158.2+3.1
GDPval Extended (Mercor held-out rubric score)Professional Work58.955.9−3.0
Conceptual Reasoning IndexIndexes68.366.2−2.9
Finance Agent v2 - Market AnalysisFinance & Accounting67.269.9+2.7
Arena Alignment IndexSafety74.577.1+2.5
LMArena Text Arena EloChat & Writing14821496+2.5
MMMU-Pro Extended (Mercor held-out)Multimodal31.433.8+2.4
MineBench — Bradley-Terry RatingGames17861747−2.4
Finance Agent v2 - ComparablesFinance & Accounting44.947.2+2.3
ACCoRDGeneral QA77.475.1−2.3
LLM Creative Story WritingChat & Writing0.80.6−2.3
Medical Long Context Reasoning (MLCR-AA)Health & Medicine45.643.3−2.2
Artificial Analysis Progress Index v1Indexes318324+2.1
GPQA Diamond (Artificial Analysis avg@5)General QA92.094.1+2.1
NoCoT-Bench NCRI 15.2General QA10298.3−2.0
LiveBench 2.0Agents76.874.8−2.0
Excel Modeling BenchmarkFinance & Accounting69.467.4−1.9
DeepSearchQA (F1 score)Search92.290.3−1.9
AA-Omniscience IndexIndexes28.825.0−1.9
AIME 2025Math98.3100.0+1.7
DTBenchGeneral QA94.996.5+1.6
Epoch Capabilities Index (ECI)Indexes160158−1.6
Terminal-Bench 4.0Code23.224.7+1.5
ProgramBenchCode1.02.5+1.5
Document ArenaProfessional Work14751473−1.5
Ruby LLM Benchmarks — Program FixerCode72.674.1+1.4
Vision Arena OverallMultimodal12861291+1.4
GBENCH One-shot Coding GScoreGames54.052.7−1.3
Bench to the Future 3 — September Panel — PooledForecasting0.10.1−0.9
SWE-TogetherCode61.562.4+0.9
OTIS Mock AIME 2024-2025Math98.399.2+0.8
VoxelBench Text RatingGames16271616−0.8
CharXiv (descriptive and reasoning combined)Multimodal93.293.9+0.7
LMArena Agent Arena Bash RecoveryAgents7.26.1−0.6
ForecastBench Preliminary Dataset Brier IndexForecasting61.661.9+0.3
ForecastBench Tournament Overall Brier IndexForecasting66.265.9−0.3
Code MigrationCode47.347.4+0.2
ObviousBenchGeneral QA99.399.30.0

Opus 4.8vsMuse Spark 1.3normalised score marginsData Nov 15, 2026Data: Vals AI, Artificial Analysis, Mercor and 39 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…