Atlas

Compare Models

Muse Ember vs Muse Wonder

Muse Wonder leads by 1.3 capability points (≈95% interval −1.1 to +3.7): statistically tied · P(higher) ≈ 85%

Wins 51 of 65 shared benchmarks (14 lost)

Capability vs. relative cost
Muse EmberMuse WonderClosed weightsOpen weights

Benchmark wins14 – 51 · 65 shared

Comparing matching setups only (1 benchmark excluded) · suites counted once
← Muse EmberMuse Wonder →
Data: Artificial Analysis, Vals AI, Surge AI and 11 othersatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Muse Wonder minus Muse Ember.

Code Migration v1.1 (Vals Index v2.2 subset)Software › Software engineering47.169.8+22.7
TaxCalcBench TY25Professional › Finance, accounting & tax2.024.0+22.0
DAXBench v2Software › Data & ML62.884.0+21.3
DAYJOB: Finance (October 2026 rescoring)Professional › Finance, accounting & tax24.238.8+14.6
DAYJOB: Healthcare (October 2026 rescoring)Health & medicine17.231.2+14.0
Humanity's Last Exam DiamondKnowledge & reasoning › General knowledge25.139.0+13.9
BuseyBench SVGPerception & media › Generation5.46.6+12.0
Terminal-Bench 4.0Software › Software engineering39.449.5+10.1
ComplexConstraints (November 2026 rescoring)Language & writing › Instruction following58.867.9+9.1
CritPt v2 (agentic, avg@5)Science › Physical sciences62.371.1+8.9
SWE Atlas v1.1 - OverallSoftware › Software engineering56.064.8+8.8
AA-Omniscience v1.1 IndexKnowledge & reasoning › General knowledge7.725.1+8.7
Artificial Analysis Economics IndexComposite indexes41.049.5+8.6
Artificial Analysis Legal IndexProfessional › Legal38.546.9+8.4
GDP.pdf (Artificial Analysis, All-pass)Professional › General knowledge work22.830.6+7.8
Artificial Analysis Engineering IndexScience › Engineering41.849.6+7.8
DeepFinanceProfessional › Finance, accounting & tax36.544.2+7.7
AA-Briefcase v1.1 EloProfessional › General knowledge work15201659+7.6
Artificial Analysis Intelligence Index v5.0.1Composite indexes36.043.6+7.6
Artificial Analysis Healthcare & Medical IndexHealth & medicine41.348.7+7.4
Ramp SWE-Bench v2Software › Software engineering51.858.6+6.8
Artificial Analysis Finance & Accounting IndexProfessional › Finance, accounting & tax44.651.1+6.6
GDP.xlsxProfessional › General knowledge work15.421.8+6.4
Artificial Analysis Coding Agent Index v2.0Software › Software engineering45.952.0+6.0
Artificial Analysis Strategy & Ops IndexProfessional › General knowledge work48.754.5+5.8
Terminal-Bench-Science 0.1Science › Multidisciplinary20.025.7+5.7
HiL-Bench (Human-in-Loop Benchmark)Tool & computer use › General agents74.079.7+5.7
RiemannBench (November 2026 rescoring)Math53.659.2+5.6
GDP.pdf (November 2026 rescoring)Professional › General knowledge work49.254.8+5.6
LMArena Code Arena WebDevSoftware › Frontend & design16081648+5.3
FrontierCUA v1 pass@1Tool & computer use › Computer & browser use7.512.5+5.0
Artificial Analysis Coding Agent Index v2.1Software › Software engineering39.744.5+4.8
APEX-1 (Mercor, mean score)Professional › General knowledge work70.475.1+4.7
Vals Index v2.2Composite indexes60.564.9+4.4
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & medicine18.322.2+3.9
OSWorld 2.1 full - reported partial reward, step limit unspecifiedTool & computer use › Computer & browser use69.572.6+3.1
Professional Reasoning Benchmark - LegalProfessional › Legal55.658.4+2.8
Professional Reasoning Benchmark - FinanceProfessional › Finance, accounting & tax56.058.7+2.7
Terminal-Bench 4.0 (Artificial Analysis)Software › Software engineering40.938.4−2.5
Artificial Analysis Progress Index v1Composite indexes329336+2.5
LMArena Agent Arena Confirmed SuccessTool & computer use › General agents-2.02.5+2.2
GBENCH One-shot Coding GScoreGames56.458.5+2.1
FORTRESS Average Risk Score (ARS)Safety & alignment11.913.8−1.8
Finance Agent v2.1Professional › Finance, accounting & tax63.865.5+1.7
APEX-SWESoftware › Software engineering52.954.6+1.7
LMArena Agent Arena Steering BurdenTool & computer use › General agents4.81.5−1.6
LMArena Agent Arena Bash ReliabilityTool & computer use › General agents6.63.4−1.6
Legal Research BenchProfessional › Legal49.050.5+1.4
τ³-Banking (Mercor run, bm25_grep retrieval)Tool & computer use › Customer service & operations53.351.9−1.4
APEX-Agents 1.1Professional › General knowledge work75.777.0+1.3
HLAB-Vals-1.1Professional › Legal25.023.8−1.2
LMArena Agent Arena Net Improvement (Steering Burden)Tool & computer use › General agents1.00.6−0.7
FrontierCode 2.0 (Artificial Analysis)Software › Software engineering27.227.9+0.7
Vibe Code Bench v1.1Software › Frontend & design83.783.1−0.6
EnterpriseBench: CoreCraftTool & computer use › Customer service & operations62.161.5−0.6
LMArena Agent Arena Tool HallucinationTool & computer use › Tool calling0.9-0.2−0.5
OSWorld 2.1 full - reported binary completion, step limit unspecifiedTool & computer use › Computer & browser use33.834.3+0.5
Vals Legal IndexProfessional › Legal75.775.3−0.4
Excel Modeling Benchmark v1.1Professional › Finance, accounting & tax72.272.5+0.3
Vals Finance IndexProfessional › Finance, accounting & tax60.961.2+0.3
Harvey LAB (Mercor run, undisclosed ~213-task cohort)Professional › Legal22.723.0+0.3
HANDBOOK.md Agents v1.1Tool & computer use › Customer service & operations29.829.5−0.3
Tax Agent BenchProfessional › Finance, accounting & tax71.471.2−0.2
LMArena Agent Arena Praise vs ComplaintTool & computer use › General agents-2.7-3.0−0.1
AutomationBench-AA 1.0.6Tool & computer use › Customer service & operations68.168.10.0
Data: Artificial Analysis, Vals AI, Surge AI and 11 othersatlas.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…