Atlas

Compare Models

Opus 4.8 vs Muse Wonder

Muse Wonder leads by 3.8 capability points (≈95% interval +1.7 to +5.9): a clear lead over Opus 4.8 · P(higher) ≈ 100%

Wins 82 of 101 shared benchmarks (18 lost, 1 tied)

Capability vs. relative cost
Opus 4.8Muse WonderClosed 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 Wondercapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins18 – 82 · 101 shared, 1 tied

Comparing matching setups only (33 benchmarks excluded) · suites counted once
← Opus 4.8Muse Wonder →

Opus 4.8vsMuse Wonderwins by categoryData Nov 15, 2026Data: Vals AI, Center for AI Safety, Mercor and 27 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Muse Wonder minus Opus 4.8.

LisanBench Difficulty-WeightedGames26948303+60.1
LisanBench Path LengthGames946327973+57.5
HiL-Bench (Human-in-Loop Benchmark)Agents35.379.7+44.3
BullshitBench v2 LeaderboardSafety94.057.0−37.0
DeepResearch Bench II (Mercor 130-task set)Search44.610.8−33.8
Terminal-Bench 4.0Code23.249.5+26.3
MindCube TinyMultimodal64.991.1+26.2
DeepSWE v1.1Code54.478.8+24.4
AutomationBench-AA 1.0.6Agents45.668.1+22.5
PCT Political Inconsistency (CAIS)Safety53.833.3+20.5
ChessBench CoherenceGames79.7100.0+20.3
Humanity's Sixth SenseMultimodal23.543.7+20.2
VCT-Refusal (CAIS)Safety41.260.2+19.0
Terminal-Bench Science 0.1 (Vals)Science4.322.9+18.6
AA-Briefcase v1.1 EloAgents13201659+18.5
SpatialViz-BenchMultimodal65.283.5+18.3
VoxelBench Text RatingGames16271870+17.3
Terminal-Bench 4.0 (Artificial Analysis)Code21.738.4+16.7
Humanity's Last Exam DiamondGeneral QA23.139.0+15.9
Bug Hunt BenchCode14.330.2+15.9
ERQAMultimodal59.575.3+15.8
EnigmaEvalGeneral QA20.536.1+15.6
GameBench 2Games62.577.8+15.4
Code MigrationCode47.362.6+15.3
IntPhys 2Multimodal55.170.2+15.1
Mystery Game PuzzlesGames36.051.0+15.0
FrontierMath Tier 4 (v2) - PrivateMath56.170.7+14.6
DistressBenchSafety71.485.9+14.5
BuseyBench SVGMultimodal5.26.6+13.9
GDPval-AA v2.1Professional Work14611770+13.6
LLM Creative Story WritingChat & Writing0.82.0+13.6
SpeciEval (Pooled English Runs)Safety84.397.5+13.2
APEX-Agents 1.1Professional Work64.477.0+12.6
APEX-1 (Mercor, mean score)Professional Work62.775.1+12.4
LMArena Code Arena WebDevCode15551649+12.2
HPCT Refusal (CAIS)Safety85.097.0+12.0
MMMU-Pro Extended (Mercor held-out)Multimodal31.443.2+11.8
FrontierCode 2.0 - Pass RateCode17.128.8+11.7
Furniture Assembly BenchmarkMultimodal42.554.2+11.7
AA-LCR Extended (Mercor held-out)Search63.274.4+11.2
GDPval Extended (Mercor held-out rubric score)Professional Work58.970.0+11.1
APEX-SWECode43.954.6+10.7
Artificial Analysis Intelligence Index v4.3.2Indexes41.852.4+10.6
FrontierCode 2.0 - ScoreCode15.325.3+10.0
MASKSafety94.484.9−9.5
EnterpriseBench: CoreCraftProfessional Work52.361.5+9.2
OncologyBench Strict (25 tasks, October 9 rubric)Health & Medicine16.024.0+8.0
GDP.pdf (Artificial Analysis, All-pass)Professional Work22.830.6+7.8
TextQuests Game ProgressGames40.147.6+7.5
Harvey's Legal Agent BenchmarkLegal9.617.1+7.5
Legal Research BenchLegal43.850.5+6.7
Finance Agent v2 - PrecedentsFinance & Accounting25.632.1+6.5
Tax Agent BenchFinance & Accounting64.971.2+6.3
Artificial Analysis Progress Index v1Indexes318336+6.1
SpatialBench V1 Clean RerunsScience65.871.9+6.1
Finance Agent v2 - General Quantitative AnalysisFinance & Accounting71.677.6+6.0
Vals Index v2.1Indexes55.161.1+6.0
SciCode 1.0.1 (Artificial Analysis)Code54.460.3+5.9
NoCoT-Bench NCKI (kspine v3)General QA117109−5.4
Finance Agent v2 - Market AnalysisFinance & Accounting67.262.1−5.1
LMArena Text Arena EloChat & Writing14821511+5.1
ChessBench EloGames9611046+5.1
MineBench — Bradley-Terry RatingGames17861868+5.0
HealthBench Professional (Length-Adjusted)Health & Medicine55.860.5+4.7
Finance Agent v2 - Earnings AnalysisFinance & Accounting72.176.8+4.7
Finance Agent v2 - General Qualitative AnalysisFinance & Accounting66.470.8+4.4
Finance Agent v2 - AdjustmentsFinance & Accounting48.452.7+4.3
GBENCH One-shot Coding GScoreGames54.058.1+4.0
BioTIER RefusalSafety94.798.7+4.0
HardGeoBench Avg DistanceMultimodal12451369−3.8
Finance Agent v2 - ComparablesFinance & Accounting44.947.9+3.1
CharXiv Extended (Mercor held-out)Multimodal76.779.6+2.9
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine19.422.2+2.8
GDPval (Mercor 213-task rubric score)Professional Work84.587.0+2.5
HardGeoBench <=100miMultimodal36.338.8+2.5
Finance Agent v2 - All PassFinance & Accounting43.545.9+2.4
ChessBench AccuracyGames72.470.2−2.4
NoCoT-Bench NCRI 15.2General QA10297.7−2.3
Excel Modeling BenchmarkFinance & Accounting69.467.1−2.3
Humanity's Last ExamGeneral QA49.847.6−2.2
FrontierMath Tiers 1-3 (v2) - PrivateMath80.082.1+2.1
Public Benefits Bench v1.1Professional Work68.170.2+2.1
TaxCalcBench TY25Finance & Accounting20.022.0+2.0
Epoch Capabilities Index (ECI)Indexes160162+1.7
AA-Omniscience IndexIndexes28.825.5−1.6
Finance Agent v2 - Disclosure AnalysisFinance & Accounting65.767.3+1.5
LiveBench 2.0Agents76.878.1+1.3
CharXiv (descriptive and reasoning combined)Multimodal93.294.5+1.3
HardGeoBench Country AccuracyMultimodal53.852.5−1.2
ACCoRDGeneral QA77.476.2−1.2
OTIS Mock AIME 2024-2025Math98.399.4+1.1
SciCode (main problems)Science16.915.8−1.1
Finance Agent v2 - Financial ModelingFinance & Accounting23.322.5−0.8
TextQuests Average Cumulative HarmSafety21.020.8+0.8
Conceptual Reasoning IndexIndexes68.367.7−0.8
LMCAGeneral QA57.556.8−0.7
Machiavelli Deception (CAIS)Safety83.783.6+0.5
MMLU ProGeneral QA90.089.6−0.4
Vibe Code Bench v1.1Code82.783.1+0.4
DTBenchGeneral QA94.995.2+0.3
HardGeoBench <=10miMultimodal27.527.50.0

Opus 4.8vsMuse Wondernormalised score marginsData Nov 15, 2026Data: Vals AI, Center for AI Safety, Mercor and 27 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…