Atlas

Compare Models

GPT-5.5 vs Muse Wonder

Muse Wonder leads by 3.5 capability points (≈95% interval +1.4 to +5.7): a clear lead over GPT-5.5 · P(higher) ≈ 100%

Wins 67 of 95 shared benchmarks (27 lost, 1 tied)

Capability vs. relative cost
GPT-5.5Muse 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.

GPT-5.5vsMuse Wondercapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins27 – 67 · 95 shared, 1 tied

Comparing matching setups only (46 benchmarks excluded) · suites counted once
← GPT-5.5Muse Wonder →

GPT-5.5vsMuse Wonderwins by categoryData Nov 15, 2026Data: Vals AI, Center for AI Safety, Mercor and 24 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Muse Wonder minus GPT-5.5.

HPCT Refusal (CAIS)Safety29.097.0+68.0
LisanBench Path LengthGames984527973+56.3
LisanBench Difficulty-WeightedGames33168303+53.4
VCT-Refusal (CAIS)Safety16.060.2+44.2
HiL-Bench (Human-in-Loop Benchmark)Agents39.779.7+40.0
DeepResearch Bench II (Mercor 130-task set)Search45.410.8−34.6
Terminal-Bench 4.0 (Artificial Analysis)Code9.138.4+29.3
AA-Briefcase v1.1 EloAgents10871610+28.6
BioTIER RefusalSafety73.498.7+25.4
AutomationBench-AA 1.0.6Agents47.368.1+20.8
GDPval-AA v2.1Professional Work13571770+18.2
ChessBench EloGames13491046−18.0
APEX-SWECode36.654.6+18.0
GameBench 2Games59.977.8+17.9
Code MigrationCode45.262.6+17.4
GDPval Extended (Mercor held-out rubric score)Professional Work54.570.0+15.5
DeepSWE v1.1Code64.478.8+14.4
Artificial Analysis Intelligence Index v4.3.2Indexes38.452.4+14.0
Harvey's Legal Agent BenchmarkLegal3.817.1+13.3
Vibe Code Bench v1.1Code69.883.1+13.2
Vending-Bench 2Games75249432+11.6
MindCube TinyMultimodal79.991.1+11.2
Humanity's Last Exam DiamondGeneral QA27.939.0+11.1
APEX-Accounting 1.1Finance & Accounting56.367.1+10.8
Tax Agent BenchFinance & Accounting60.571.2+10.7
IntPhys 2Multimodal59.970.2+10.3
Finance Agent v2 - PrecedentsFinance & Accounting21.932.1+10.2
EnterpriseBench: CoreCraftProfessional Work51.361.5+10.2
Legal Research BenchLegal40.450.5+10.1
Furniture Assembly BenchmarkMultimodal44.254.2+10.0
BullshitBench v2 LeaderboardSafety47.057.0+10.0
LiveBench 2.0Agents68.778.1+9.4
Public Benefits Bench v1.1Professional Work60.970.2+9.3
OncologyBench Strict (25 tasks, October 9 rubric)Health & Medicine14.724.0+9.3
SpatialViz-BenchMultimodal74.283.5+9.3
ChessBench AccuracyGames78.870.2−9.2
Finance Agent v2 - Disclosure AnalysisFinance & Accounting58.967.3+8.3
Creative Writing v3 EloChat & Writing18452007+8.2
GDP.pdf (Artificial Analysis, All-pass)Professional Work22.630.6+8.0
SpatialBench V1 Clean RerunsScience64.171.9+7.8
Finance Agent v2 - AdjustmentsFinance & Accounting45.752.7+6.9
APEX-Agents 1.1Professional Work70.377.0+6.7
PCT Political Inconsistency (CAIS)Safety40.033.3+6.7
APEX-1 (Mercor, mean score)Professional Work68.475.1+6.7
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine28.922.2−6.7
Finance Agent v2 - All PassFinance & Accounting39.645.9+6.4
BuseyBench SVGMultimodal6.06.6+6.0
Machiavelli Deception (CAIS)Safety81.783.6−5.7
Artificial Analysis Progress Index v1Indexes320336+5.5
LLM Creative Story WritingChat & Writing2.42.0−5.3
Finance Agent v2 - General Qualitative AnalysisFinance & Accounting65.670.8+5.2
MASKSafety90.184.9−5.2
LMArena Text Arena EloChat & Writing14811511+5.2
Mystery Game PuzzlesGames56.051.0−5.0
ERQAMultimodal70.575.3+4.8
ACCoRDGeneral QA80.576.2−4.3
Finance Agent v2 - ComparablesFinance & Accounting43.747.9+4.3
DAXBench v2Code88.384.0−4.3
GBENCH One-shot Coding GScoreGames62.258.1−4.2
SciCode 1.0.1 (Artificial Analysis)Code56.160.3+4.2
VoxelBench Text RatingGames19251870−3.9
AA-LCR Extended (Mercor held-out)Search70.574.4+3.9
Finance Agent v2 - General Quantitative AnalysisFinance & Accounting73.777.6+3.9
Humanity's Last ExamGeneral QA43.847.6+3.8
MMMU ProMultimodal81.284.9+3.7
Finance Agent v2 - Earnings AnalysisFinance & Accounting73.276.8+3.6
SciCode (main problems)Science12.315.8+3.5
FrontierMath Tiers 1-3 (v2) - PrivateMath85.382.1−3.2
CharXiv Extended (Mercor held-out)Multimodal82.779.6−3.1
CritPt (avg@5)Science25.428.3+2.9
LMCAGeneral QA54.356.8+2.6
Excel Modeling BenchmarkFinance & Accounting64.567.1+2.5
AA-Omniscience IndexIndexes20.525.5+2.5
GDPval (Mercor 213-task rubric score)Professional Work84.687.0+2.4
TextQuests Average Cumulative HarmSafety21.420.8+2.2
TaxCalcBench TY25Finance & Accounting24.022.0−2.0
Finance Agent v2 - Market AnalysisFinance & Accounting60.362.1+1.8
FrontierMath Tier 4 (v2) - PrivateMath72.570.7−1.8
Creative Writing v3 Rubric ScoreChat & Writing15.916.2+1.6
Epoch Capabilities Index (ECI)Indexes160162+1.6
NoCoT-Bench NCRI 15.2General QA10097.7−1.4
MineBench — Bradley-Terry RatingGames18461868+1.3
TextQuests Game ProgressGames48.947.6−1.2
EnigmaEvalGeneral QA37.236.1−1.1
NoCoT-Bench NCKI (kspine v3)General QA111109−0.9
DTBenchGeneral QA96.095.2−0.8
MMLU ProGeneral QA88.989.6+0.7
SpeciEval (Pooled English Runs)Safety98.297.5−0.7
MMMU-Pro Extended (Mercor held-out)Multimodal43.843.2−0.6
Conceptual Reasoning IndexIndexes67.367.7+0.6
OTIS Mock AIME 2024-2025Math100.099.4−0.6
Finance Agent v2 - Financial ModelingFinance & Accounting22.922.5−0.4
Code2Video-AA EloCode1000997−0.2
CharXiv (descriptive and reasoning combined)Multimodal94.694.5−0.1
ChessBench CoherenceGames100.0100.00.0

GPT-5.5vsMuse Wondernormalised score marginsData Nov 15, 2026Data: Vals AI, Center for AI Safety, Mercor and 24 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…