Atlas

Compare Models

Gemini 3.8 Flash vs Muse Spark 1.3

Muse Spark 1.3 leads by 0.5 capability points (≈95% interval −1.6 to +2.6): statistically tied · P(higher) ≈ 68%

Wins 139 of 232 shared benchmarks (91 lost, 2 tied)

Capability vs. relative cost
Gemini 3.8 FlashMuse 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.

Gemini 3.8 FlashvsMuse Spark 1.3capability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins91 – 139 · 232 shared, 2 tied

Comparing matching setups only (18 benchmarks excluded) · suites counted once
← Gemini 3.8 FlashMuse Spark 1.3 →

Gemini 3.8 FlashvsMuse Spark 1.3wins by categoryData Nov 15, 2026Data: Vals AI, Artificial Analysis, Mercor and 60 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 Gemini 3.8 Flash.

AA-PersonalityBench Imagination indexChat & Writing41.098.1+57.7
SAFE-Teen 106 worst branch: Health Advice (no prompt)Safety69.013.9+55.1
BabaIsBench Stage 0: The IntroGames100.046.0−54.0
SAFE-Teen 106: Health Advice (no prompt)Safety59.66.7+52.9
CyberGym-E2E-AA (Artificial Analysis)Cybersecurity8.459.5+51.1
SAFE-Teen 106 worst branch: Health Advice (under-18 prompt)Safety48.14.5+43.6
Design Creativity BenchCode60.647.2−42.6
AA-PersonalityBench Warmth indexChat & Writing60.096.2+36.5
SAFE-Teen 106: Health Advice (under-18 prompt)Safety37.52.1+35.5
BrokenArXiv 06/2026Math20.853.2+32.4
SAFE-Teen 106 worst branch: Urgent Safety (no prompt)Safety56.425.6+30.8
CyberBench v1.1 OverallCybersecurity43.872.7+29.0
HiL-Bench (Human-in-Loop Benchmark)Agents41.566.9+25.4
WordleBenchGames98.073.0−25.0
ClockBench AccuracyMultimodal66.742.2−24.5
FrontierMath Tier 4 (v2) - PrivateMath22.046.3+24.4
SAFE-Teen 106 worst branch: Urgent Safety (under-18 prompt)Safety34.811.3+23.5
RiemannBenchMath51.228.0−23.2
AA-Briefcase EloAgents12021589+22.5
SAFE-Teen 106: Urgent Safety (no prompt)Safety31.59.0+22.4
Mystery Game PuzzlesGames47.025.0−22.0
Medical Long Context Reasoning (MLCR-AA)Health & Medicine21.743.3+21.7
Last Translation Benchmark (LTBv1-eval, blind)Chat & Writing39.818.4−21.4
RuneBenchGames6.04.6−21.1
TextQuests Game ProgressGames54.133.0−21.1
Chess Puzzles (task version 2.0)Games58.037.0−21.0
RSI Arena Stage 1 human ratingAgents981938−20.7
ChessBench EloGames15271201−19.4
ITBench-AAAgents52.533.2−19.3
AA-PersonalityBench Informativeness indexChat & Writing81.0100+19.2
Artificial Analysis Cyber Index v1Indexes25.144.3+19.2
MA-BenchProfessional Work43.661.8+18.2
BrokenArXiv 05/2026Math19.036.5+17.5
WorldAuditBenchMultimodal32.415.0−17.4
AA-PersonalityBench Transparency indexChat & Writing23.841.0+17.3
Legal Research BenchLegal38.955.3+16.3
Olam Social Poker EloGames14791494+16.3
Shroombench (mushroom identification, top-1)Multimodal65.049.0−16.0
BrokenArXiv 04/2026Math33.649.2+15.6
Finance Agent v2 - General Qualitative AnalysisFinance & Accounting68.183.4+15.3
NoCoT-Bench NCRI 15.2General QA12698.3−15.1
ArXivMath 05/2026Math58.373.3+15.0
GeneBench-Pro Public Release SubsetScience11.026.0+15.0
DistressBenchSafety70.784.9+14.1
WellnessBenchSafety47.061.0+14.0
Harvey's Legal Agent BenchmarkLegal10.023.8+13.8
Terminal-Bench 4.0 (Artificial Analysis)Code19.733.3+13.6
Bug Hunt BenchCode17.130.7+13.5
ChartographyMultimodal40.927.6−13.3
MentalHealthBenchSafety35.548.6+13.1
Professional Reasoning Benchmark - LegalLegal48.961.6+12.7
Artificial Analysis Coding Agent Index v1.5Indexes41.954.3+12.4
Artificial Analysis Coding Agent Index v2.0Indexes33.345.7+12.4
Extended NYT Connections 940 (Quadratic v1)Games97.485.1−12.3
SAFE-Teen 106: Urgent Safety (under-18 prompt)Safety14.92.7+12.2
DAYJOB: FinanceFinance & Accounting2.814.8+12.0
CadQueryEval (Corrected Geometry)Code96.084.0−12.0
DAYJOB: AccountingFinance & Accounting6.018.0+12.0
AA-LCR Extended (Mercor held-out)Search55.666.9+11.3
SnorkelLegalLegal25.336.5+11.2
LLM Debate BenchmarkChat & Writing15141593+11.2
Visa Agentic Commerce BenchmarkAgents53.364.4+11.1
Code MigrationCode36.547.4+10.9
SAFE-Teen 106 worst branch: Relationships (under-18 prompt)Safety14.03.2+10.7
GDPval-AA v2.1Professional Work14421681+10.5
AutoResearchExam (hidden-test AUARC)Agents0.50.4−10.5
Gray Swan IPI Attack Success (k=15)Safety5.515.9−10.4
FrontierML ($200 wallet)Agents5.716.0+10.3
Professional Reasoning Benchmark - FinanceFinance & Accounting49.559.5+10.0
ProofBench v1.1Math48.058.0+10.0
GDPval-AA v2Professional Work14641662+9.8
Code2Video-AA EloCode8991019+9.7
AA-PersonalityBench Organization indexChat & Writing73.363.8−9.6
Arena Image-to-WebDevCode15941637+9.5
LMArena Code Arena WebDevCode15831655+9.5
ArXivMath 06/2026Math67.476.7+9.4
BuseyBench SVGMultimodal6.65.7−9.2
CAD ArenaAgents41.031.8−9.2
AutomationBench 1.0.6 (Zapier)Agents29.720.7−9.0
SpeciEval (Refusal-Aware Retries, Historical)Safety89.998.8+8.9
Harvey LAB-AA v1.1 (hallucination-gated)Legal0.08.9+8.9
SpeciEval (Pooled English Runs)Safety89.998.7+8.9
AA-AnalystAgent (pass^5)Professional Work46.337.5−8.7
ArXivLean 06/2026Math19.628.3+8.7
Finance Agent v2 - Market AnalysisFinance & Accounting78.569.9−8.6
MedXpertQA MM (200-question set)Health & Medicine84.776.3−8.4
AA-3D v1.0 EloCode9741032+8.3
CWE-Bench-AA (Artificial Analysis)Cybersecurity52.560.8+8.3
micro1 PharmaBenchHealth & Medicine5.613.9+8.3
Creative Writing v3 EloChat & Writing17441907+8.2
Document ArenaProfessional Work14821473−8.1
CritPt (avg@5)Science18.326.0+7.7
AA-PersonalityBench Conviction indexChat & Writing84.877.2−7.7
AA-PersonalityBench Expressiveness indexChat & Writing79.171.4−7.7
Snorkel Agentic Coding 2.0Code32.625.0−7.6
HardGeoBench <=10miMultimodal46.338.8−7.5
HardGeoBench Country AccuracyMultimodal67.560.0−7.5
LiveBench 2.0Agents67.374.8+7.4
Vibe Code Bench v1.1Code78.785.9+7.2
SnorkelRevOpsProfessional Work5.412.6+7.2
Artificial Analysis Intelligence Index v4.3.2Indexes40.948.1+7.2
BrokenArXiv 08/2026 (Native Harnesses)Math27.420.2−7.1
ALE-BenchCode21821981−7.1
Decision-Bench v1.1Professional Work26.919.8−7.1
DAYJOB: HealthcareHealth & Medicine0.87.8+7.0
Artificial Analysis Intelligence Index v4.3Indexes41.248.2+7.0
Tax Agent BenchFinance & Accounting66.873.4+6.6
SAFE-Teen 106 worst branch: Relationships (no prompt)Safety62.656.2+6.4
FrontierSWE V2Code19.625.8+6.2
Longform Creative WritingChat & Writing76.882.8+6.0
Realm Financial ReasoningFinance & Accounting51.545.5−6.0
FrontierMath Tiers 1-3 (v2) - PrivateMath68.474.4+6.0
GBENCH One-shot Coding GScoreGames58.652.7−5.9
Artificial Analysis Intelligence Index v5.0Indexes30.536.4+5.9
Artificial Analysis Intelligence Index v5.0.1Indexes30.836.6+5.8
SAFE-Teen 106: Relationships (no prompt)Safety40.134.4+5.8
Terminal-Bench Science 0.1 (Vals)Science8.614.3+5.7
PCT Political Inconsistency (CAIS)Safety46.340.6+5.6
HardGeoBench Avg DistanceMultimodal8431022−5.5
NoCoT-Bench NCKI (kspine v3)General QA121113−5.5
ChessBench AccuracyGames81.176.0−5.5
Roboflow Vision Evals (September 2026 protocol)Multimodal85.179.8−5.3
ArXivMath 08/2026 (Native Harnesses)Math40.445.6+5.3
AURORA Multilingual Terminal-BenchCode67.762.5−5.2
Access to Justice BenchmarkLegal67.172.2+5.1
APEX-1 (Mercor, mean score)Professional Work68.873.9+5.1
Terminal-Bench 4.0Code19.724.7+5.1
RSI-Exam 0.1 (Full 88)Agents0.30.4+5.0
HardGeoBench <=100miMultimodal60.055.0−5.0
Finance Agent v2 - ComparablesFinance & Accounting52.047.2−4.8
Excel Modeling BenchmarkFinance & Accounting72.267.4−4.8
𝜏³-BankingAgents45.850.5+4.7
Financial Coaching BenchFinance & Accounting66.371.0+4.7
LLM Creative Story WritingChat & Writing0.10.6+4.6
HANDBOOK.mdProfessional Work10.815.4+4.6
SnorkelFinance 2.0Finance & Accounting5.410.0+4.6
Humanity's Last Exam DiamondGeneral QA29.133.7+4.6
GameBench 2Games64.068.5+4.5
GDPval Extended (Mercor held-out rubric score)Professional Work60.455.9−4.5
Westworld Finance Diligence BenchFinance & Accounting49.745.2−4.5
GDP.pdfProfessional Work23.227.6+4.4
FrontierCode 2.0 (Artificial Analysis)Code18.723.1+4.4
Hedge-BenchFinance & Accounting4.79.1+4.4
GDP.xlsxProfessional Work18.322.6+4.3
Finance Agent v2 - Earnings AnalysisFinance & Accounting81.477.2−4.2
Humanity's Sixth SenseMultimodal41.637.4−4.2
Snorkel SWE-bench CLICode6.32.2−4.1
GeoBenchMultimodal43074103−4.1
Eval Connections One-Shot with TrapsGames88.084.0−4.0
LMArena Agent Arena Bash ReliabilityAgents9.92.0−3.9
AA-PersonalityBench Slop indexChat & Writing71.475.3+3.8
Harvey LAB (Mercor run, undisclosed ~213-task cohort)Legal11.415.0+3.6
CVDP (517-problem set)Code50.754.3+3.6
MineBench — Bradley-Terry RatingGames18061747−3.6
FrontierCode 1.1 Main (Artificial Analysis)Code37.140.6+3.5
ComplexConstraintsChat & Writing48.451.9+3.5
SnorkelManufacturingProfessional Work7.110.5+3.4
EnterpriseBench: CoreCraftProfessional Work58.561.9+3.4
ChessBench CoherenceGames94.691.2−3.4
LMArena Agent Arena Bash RecoveryAgents-0.76.1+3.4
AIME 2025Math96.7100.0+3.3
Vals Index v2.1Indexes54.958.2+3.3
Ramp Accounting BenchFinance & Accounting45.742.5−3.3
SciCode 1.0.1 (Artificial Analysis)Code56.659.7+3.1
SAFE-Teen 106: Relationships (under-18 prompt)Safety3.70.6+3.1
HoneyBench v0.2 (hack rate)Safety80.583.6−3.1
ACCoRDGeneral QA78.275.1−3.1
AutomationBench-AA 1.0.6Agents60.957.9−3.0
Humanity's Last ExamGeneral QA46.243.2−3.0
CharXiv Extended (Mercor held-out)Multimodal78.981.8+2.9
CritPt v2 (agentic, avg@5)Science65.168.0+2.9
Finance Agent v2 - General Quantitative AnalysisFinance & Accounting80.978.0−2.9
SheetBench-50Finance & Accounting45.642.8−2.8
DAYJOB: TaxFinance & Accounting0.83.6+2.8
APEX-SWECode36.439.1+2.7
MMMU-Pro Extended (Mercor held-out)Multimodal36.433.8−2.6
APEX-Agents 1.1Professional Work72.970.4−2.5
ArXivMath 04/2026Math64.266.7+2.5
Real-SWECode38.836.3−2.5
SWE-bench Multimodal (Vals v1, 480 Tasks)Code40.843.1+2.3
AA-Omniscience IndexIndexes29.625.0−2.3
Vals Index v2Indexes62.364.5+2.3
MatFlowBench Pass@1Science18.916.7−2.2
sudo L7Code11.713.9+2.2
DAXBench v2Code87.285.1−2.1
Finance Agent v2 - AdjustmentsFinance & Accounting54.852.8−2.0
DeepsecBench-AA (Artificial Analysis)Cybersecurity14.512.5−2.0
LMArena Agent Arena Steering BurdenAgents-1.32.6+1.9
SWE-TogetherCode64.262.4−1.8
APEX-AccountingFinance & Accounting51.753.5+1.8
CUA-benchAgents4.25.8+1.7
LMArena Agent Arena Praise vs ComplaintAgents4.91.7−1.6
Finance Agent v2 - Disclosure AnalysisFinance & Accounting68.570.1+1.6
Immigration BenchLegal63.465.0+1.6
DeepSWE v1.1Code73.875.4+1.6
CharXiv (descriptive and reasoning combined)Multimodal95.493.9−1.5
ProgramBenchCode1.02.5+1.5
Artificial Analysis Progress Index v1Indexes320324+1.5
ForecastBench Preliminary Dataset Brier IndexForecasting60.561.9+1.4
LMArena Agent Arena SteerabilityAgents4.31.6−1.3
Arena Alignment IndexSafety75.877.1+1.3
LMArena Agent Arena Confirmed SuccessAgents1.33.7+1.2
Terminal-Bench 2.1Code87.688.8+1.2
GPQA Diamond (Artificial Analysis avg@5)General QA95.394.1−1.1
Vals RSI Index v1.1 (historical five tasks)Agents17.618.7+1.1
AA-LCR v1.1Search84.083.0−1.0
LMCAGeneral QA52.953.9+1.0
Finance Agent v2 - PrecedentsFinance & Accounting36.135.2−0.9
Finance Agent v2 - Financial ModelingFinance & Accounting32.633.5+0.9
DTBenchGeneral QA95.796.5+0.8
LegalReasonBenchLegal39.640.4+0.8
MedXpertQA MM Extended (Mercor held-out)Health & Medicine60.860.1−0.7
ObviousBenchGeneral QA100.099.3−0.7
Conceptual Reasoning IndexIndexes65.766.2+0.7
Snorkel Digital Work IndexIndexes16.916.3−0.6
SimpleBenchGeneral QA82.481.8−0.6
Epoch Capabilities Index (ECI)Indexes159158−0.6
SnorkelUnderwrite 2.0Finance & Accounting19.119.6+0.5
Terminal-Bench-Science 0.1Science10.511.0+0.5
MysteryMechanismScience36.536.0−0.4
Vision Arena OverallMultimodal12931291−0.4
IOI (Vals v2)Code56.956.6−0.4
Vals RSI Index v1.1 (four tasks)Agents19.919.6−0.3
OTIS Mock AIME 2024-2025Math98.999.2+0.3
LMArena Agent Arena Net ImprovementAgents3.83.8+0.2
LMArena Agent Arena Net Improvement (Steering Burden)Agents2.32.2−0.2
LMArena Text Arena EloChat & Writing14971496−0.2
Creative Writing v3 Rubric ScoreChat & Writing15.615.6+0.2
Finance Agent v2 - All PassFinance & Accounting49.749.5−0.2
LMArena Agent Arena Tool HallucinationAgents0.40.8+0.2
SWE-bench Verified Extended (Mercor held-out)Code54.054.00.0
TaxCalcBench TY25Finance & Accounting6.06.00.0

Gemini 3.8 FlashvsMuse Spark 1.3normalised score marginsData Nov 15, 2026Data: Vals AI, Artificial Analysis, Mercor and 60 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…