Atlas

Compare Models

Gemini 3.8 Flash vs Muse Wonder

Muse Wonder leads by 2.9 capability points (≈95% interval +0.6 to +5.2): a clear lead over Gemini 3.8 Flash · P(higher) ≈ 99%

Wins 138 of 180 shared benchmarks (40 lost, 2 tied)

Capability vs. relative cost
Gemini 3.8 FlashMuse 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.

Gemini 3.8 FlashvsMuse Wondercapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins40 – 138 · 180 shared, 2 tied

Comparing matching setups only (14 benchmarks excluded) · suites counted once
← Gemini 3.8 FlashMuse Wonder →

Gemini 3.8 FlashvsMuse Wonderwins by categoryData Nov 15, 2026Data: Vals AI, Artificial Analysis, Mercor and 38 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Muse Wonder minus Gemini 3.8 Flash.

HPCT Refusal (CAIS)Safety13.097.0+84.0
VCT-Refusal (CAIS)Safety6.460.2+53.8
AA-PersonalityBench Imagination indexChat & Writing41.090.5+50.0
FrontierMath Tier 4 (v2) - PrivateMath22.070.7+48.8
AA-PersonalityBench Warmth indexChat & Writing60.0100+40.4
HiL-Bench (Human-in-Loop Benchmark)Agents41.579.7+38.2
CyberBench v1.1 OverallCybersecurity43.880.4+36.6
ProofBench v1.1Math48.084.0+36.0
Code Migration v1.1Code38.672.4+33.7
AA-PersonalityBench Transparency indexChat & Writing23.856.2+32.7
Terminal-Bench 4.0Code19.749.5+29.8
Visa Agentic Commerce BenchmarkAgents53.382.2+28.9
ChessBench EloGames15271046−28.6
MASKSafety58.684.9+26.3
Vending-Bench 2Games50949432+26.3
Code MigrationCode36.562.6+26.0
DAYJOB: AccountingFinance & Accounting6.030.8+24.8
Furniture Assembly BenchmarkMultimodal31.754.2+22.5
AA-Briefcase v1.1 EloAgents12041610+22.2
Artificial Analysis Coding Agent Index v1.5Indexes41.963.7+21.8
SnorkelManufacturingProfessional Work7.127.8+20.7
LLM Creative Story WritingChat & Writing0.12.0+20.5
AA-LCR Extended (Mercor held-out)Search55.674.4+18.8
Artificial Analysis Coding Agent Index v2.0Indexes33.352.0+18.7
Terminal-Bench 4.0 (Artificial Analysis)Code19.738.4+18.7
DAYJOB: FinanceFinance & Accounting2.821.3+18.5
APEX-SWECode36.454.6+18.2
AA-PersonalityBench Slop indexChat & Writing71.454.3−17.3
AA-PersonalityBench Organization indexChat & Writing73.390.5+17.3
DAYJOB: HealthcareHealth & Medicine0.818.0+17.2
Finance Agent v2 - Market AnalysisFinance & Accounting78.562.1−16.4
TaxCalcBench TY25Finance & Accounting6.022.0+16.0
NoCoT-Bench NCRI 15.2General QA12697.7−15.4
DistressBenchSafety70.785.9+15.2
RuneBenchGames6.05.0−15.0
Eval Connections One-Shot with TrapsGames88.073.0−15.0
BioTIER RefusalSafety83.998.7+14.8
GDPval-AA v2.1Professional Work14421770+14.5
Terminal-Bench Science 0.1 (Vals)Science8.622.9+14.3
GameBench 2Games64.077.8+13.8
FrontierMath Tiers 1-3 (v2) - PrivateMath68.482.1+13.7
FrontierML ($200 wallet)Agents5.719.3+13.6
AA-PersonalityBench Informativeness indexChat & Writing81.094.3+13.5
Creative Writing v3 EloChat & Writing17442007+13.3
Bug Hunt BenchCode17.130.2+13.0
CrossSearch (pass^3)Search29.042.0+13.0
PCT Political Inconsistency (CAIS)Safety46.333.3+13.0
SnorkelLegalLegal25.338.3+13.0
MacroscopeBenchCode51.364.2+12.9
Artificial Analysis Intelligence Index v5.0Indexes30.543.3+12.8
Artificial Analysis Intelligence Index v5.0.1Indexes30.843.6+12.7
SnorkelRevOpsProfessional Work5.418.0+12.6
SpatialBench V1 Clean RerunsScience59.471.9+12.5
TextQuests Average Cumulative HarmSafety24.420.8+12.4
MKA (Artificial Analysis)Professional Work35.347.5+12.2
Terminal-Bench-Science 0.1Science10.021.9+11.9
ChessBench AccuracyGames81.170.2−11.7
HLAB-Vals-1.1Legal12.123.8+11.7
Harvey LAB (Mercor run, undisclosed ~213-task cohort)Legal11.423.0+11.6
micro1 PharmaBenchHealth & Medicine5.617.2+11.6
Legal Research BenchLegal38.950.5+11.5
Artificial Analysis Intelligence Index v4.3.2Indexes40.952.4+11.5
Snorkel Agentic Coding 2.0Code32.643.6+11.0
LiveBench 2.0Agents67.378.1+10.8
OncologyBench Strict (25 tasks, October 9 rubric)Health & Medicine13.324.0+10.7
Finance Agent v2 - Financial ModelingFinance & Accounting32.622.5−10.1
CritPt (avg@5)Science18.328.3+10.0
IOI (Vals v2)Code56.966.9+10.0
AA-PersonalityBench Expressiveness indexChat & Writing79.188.6+9.6
GDPval Extended (Mercor held-out rubric score)Professional Work60.470.0+9.6
SnorkelFinance 2.0Finance & Accounting5.415.0+9.6
Professional Reasoning Benchmark - LegalLegal48.958.4+9.5
Snorkel Digital Work IndexIndexes16.926.4+9.5
FrontierCode 1.1 Main (Artificial Analysis)Code37.146.5+9.3
Professional Reasoning Benchmark - FinanceFinance & Accounting49.558.7+9.2
MysteryMechanismScience36.527.5−9.0
Vals Index v2.2Indexes56.064.9+8.9
Machiavelli Deception (CAIS)Safety86.583.6+8.9
LMArena Code Arena WebDevCode15831649+8.6
APEX-Accounting 1.1Finance & Accounting58.867.1+8.3
WellnessBenchSafety47.055.0+8.0
Code2Video-AA EloCode899997+7.9
BioSecBench-Refusal V1 Clean RerunsSafety47.455.3+7.9
FrontierCode 2.0 (Artificial Analysis)Code18.726.6+7.9
MissionReportProfessional Work54.362.1+7.8
GDP.pdf (Artificial Analysis, All-pass)Professional Work22.830.6+7.8
NoCoT-Bench NCKI (kspine v3)General QA121109−7.8
IntPhys 2Multimodal62.570.2+7.7
Decision-Bench v1.1Professional Work26.934.6+7.7
SpeciEval (Pooled English Runs)Safety89.997.5+7.6
Financial Coaching BenchFinance & Accounting66.373.8+7.5
SWE Atlas v1.1 - OverallCode57.364.8+7.5
AutomationBench-AA 1.0.6Agents60.968.1+7.2
Harvey's Legal Agent BenchmarkLegal10.017.1+7.1
MMMU-Pro Extended (Mercor held-out)Multimodal36.443.2+6.8
Toloka Arena (v3)Agents59.065.7+6.7
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine15.622.2+6.7
TextQuests Game ProgressGames54.147.6−6.5
APEX-1 (Mercor, mean score)Professional Work68.875.1+6.3
Immigration BenchLegal63.469.6+6.2
Snorkel WorkplaceAgentsProfessional Work11.217.4+6.2
Vals Index v2.1Indexes54.961.1+6.2
Access to Justice BenchmarkLegal67.173.0+5.9
AA-PersonalityBench Conviction indexChat & Writing84.879.1−5.8
CritPt v2 (agentic, avg@5)Science65.170.9+5.7
Vals Legal IndexIndexes69.875.3+5.5
Artificial Analysis Progress Index v1Indexes320336+5.4
ChessBench CoherenceGames94.6100.0+5.4
SpatialViz-BenchMultimodal78.183.5+5.4
DAYJOB: TaxFinance & Accounting0.86.0+5.2
Excel Modeling BenchmarkFinance & Accounting72.267.1−5.1
Snorkel SWE-bench CLICode6.311.4+5.1
DeepSWE v1.1Code73.878.8+5.0
Public Benefits Bench v1.1Professional Work65.370.2+4.9
BioMysteryBench (Vals Terminus 2)Science62.267.0+4.8
SWE-bench Multimodal (Vals v1, 480 Tasks)Code40.845.6+4.8
ComplexConstraints (November 2026 rescoring)Chat & Writing63.267.9+4.7
Finance Agent v2 - Earnings AnalysisFinance & Accounting81.476.8−4.6
Vibe Code Bench v1.1Code78.783.1+4.4
Tax Agent BenchFinance & Accounting66.871.2+4.4
ERQAMultimodal79.675.3−4.3
APEX-Agents 1.1Professional Work72.977.0+4.1
Finance Agent v2 - ComparablesFinance & Accounting52.047.9−4.1
Finance Agent v2 - PrecedentsFinance & Accounting36.132.1−4.0
Mystery Game PuzzlesGames47.051.0+4.0
LMCAGeneral QA52.956.8+3.9
sudo L7Code11.715.6+3.9
SnorkelUnderwrite 2.0Finance & Accounting19.123.0+3.9
Finance Agent v2 - All PassFinance & Accounting49.745.9−3.8
MineBench — Bradley-Terry RatingGames18061868+3.8
SciCode 1.0.1 (Artificial Analysis)Code56.660.3+3.7
GDPval (Mercor 213-task rubric score)Professional Work83.587.0+3.5
LMArena Agent Arena Praise vs ComplaintAgents4.9-1.8−3.4
Vals Science IndexIndexes35.839.1+3.4
Finance Agent v2 - General Quantitative AnalysisFinance & Accounting80.977.6−3.2
DAXBench v2Code87.284.0−3.2
τ³-Banking (Mercor run, bm25_grep retrieval)Agents48.851.9+3.1
Creative Writing v3 Rubric ScoreChat & Writing15.616.2+3.1
CUA-benchAgents4.27.2+3.0
EnterpriseBench: CoreCraftProfessional Work58.561.5+3.0
HANDBOOK.md Agents v1.1Professional Work26.529.5+3.0
Humanity's Last Exam DiamondGeneral QA34.337.3+3.0
Ramp Accounting BenchFinance & Accounting45.748.7+3.0
Conceptual Reasoning IndexIndexes65.767.7+2.8
Finance Agent v2 - General Qualitative AnalysisFinance & Accounting68.170.8+2.7
Epoch Capabilities Index (ECI)Indexes159162+2.7
LMArena Agent Arena Bash ReliabilityAgents9.94.8−2.5
RiemannBench (November 2026 rescoring)Math56.859.2+2.4
LMArena Text Arena EloChat & Writing14971511+2.4
Finance Agent v2 - AdjustmentsFinance & Accounting54.852.7−2.2
Humanity's Sixth SenseMultimodal41.643.7+2.1
VIALSMultimodal29.431.5+2.1
AA-Omniscience IndexIndexes29.625.5−2.0
LMArena Agent Arena Confirmed SuccessAgents1.35.4+2.0
ACCoRDGeneral QA78.276.2−2.0
LMArena Agent Arena Steering BurdenAgents-1.32.6+1.9
SciCode (main problems)Science14.215.8+1.6
Humanity's Last ExamGeneral QA46.247.6+1.4
Finance Agent v2 - Disclosure AnalysisFinance & Accounting68.567.3−1.2
MindCube TinyMultimodal92.291.1−1.1
EnigmaEvalGeneral QA37.236.1−1.1
MMLU ProGeneral QA90.789.6−1.1
LMArena Agent Arena Net Improvement (Steering Burden)Agents2.31.8−1.0
CharXiv (descriptive and reasoning combined)Multimodal95.494.5−0.9
Vals Finance IndexIndexes60.561.2+0.8
CharXiv Extended (Mercor held-out)Multimodal78.979.6+0.7
Prophet Arena Arena Score (Agentic, OpenCode)Forecasting86.085.4−0.6
GBENCH One-shot Coding GScoreGames58.658.1−0.6
OTIS Mock AIME 2024-2025Math98.999.4+0.6
DTBenchGeneral QA95.795.2−0.5
HoneyBench v0.2 (hack rate)Safety80.581.0−0.5
LMArena Agent Arena Tool HallucinationAgents0.4-0.5−0.5
MMMU ProMultimodal84.584.9+0.4
Excel Modeling Benchmark v1.1Finance & Accounting72.972.5−0.4
BuseyBench SVGMultimodal6.66.6−0.3
Finance Agent v2.1Finance & Accounting65.865.5−0.2
Build-A-Chip-Bench (20-task index run)Agents0.10.3+0.2
Prophet Arena Arena Score (Fixed Context)Forecasting85.385.2−0.1
Dual-Arm Sim (task completion)Agents0.00.00.0
DualEntry Accounting AI BenchmarkFinance & Accounting83.283.20.0

Gemini 3.8 FlashvsMuse Wondernormalised score marginsData Nov 15, 2026Data: Vals AI, Artificial Analysis, Mercor and 38 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…