Atlas

Compare Models

DeepSeek V4.1 Flash vs Muse Spark 1.3

Muse Spark 1.3 leads by 1.8 capability points (≈95% interval −0.2 to +3.8): statistically tied · P(higher) ≈ 96%

Wins 120 of 163 shared benchmarks (42 lost, 1 tied)

Capability vs. relative cost
DeepSeek V4.1 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.

DeepSeek V4.1 FlashvsMuse Spark 1.3capability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins42 – 120 · 163 shared, 1 tied

Comparing matching setups only (6 benchmarks excluded) · suites counted once
← DeepSeek V4.1 FlashMuse Spark 1.3 →

DeepSeek V4.1 FlashvsMuse Spark 1.3wins by categoryData Nov 15, 2026Data: Vals AI, Artificial Analysis, Arena and 37 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 DeepSeek V4.1 Flash.

AA-PersonalityBench Imagination indexChat & Writing31.598.1+67.3
AA-PersonalityBench Warmth indexChat & Writing41.096.2+55.8
KernelBench Mega Normalized (RTX PRO 6000)Code48.26.2−42.1
AA-PersonalityBench Slop indexChat & Writing37.275.3+38.5
CyberGym-E2E-AA (Artificial Analysis)Cybersecurity22.959.5+36.6
AA-PersonalityBench Organization indexChat & Writing10063.8−36.5
BrokenArXiv 08/2026 (Native Harnesses)Math54.820.2−34.5
SAFE-Teen 106 worst branch: Health Advice (under-18 prompt)Safety37.04.5+32.5
SWE-bench Verified Extended (Mercor held-out)Code82.754.0−28.7
SAFE-Teen 106 worst branch: Health Advice (no prompt)Safety41.113.9+27.2
RSI Arena Stage 1 human ratingAgents887938+24.5
SAFE-Teen 106: Health Advice (no prompt)Safety30.26.7+23.5
ArXivMath 04/2026Math45.066.7+21.7
SAFE-Teen 106: Health Advice (under-18 prompt)Safety23.62.1+21.5
AA-PersonalityBench Informativeness indexChat & Writing79.1100+21.2
Medical Long Context Reasoning (MLCR-AA)Health & Medicine22.843.3+20.6
EnterpriseBench: CoreCraftProfessional Work41.561.9+20.4
FrontierMath Tier 4 (v2) - PrivateMath26.846.3+19.5
AA-PersonalityBench Conviction indexChat & Writing96.277.2−19.2
AA-PersonalityBench Transparency indexChat & Writing60.041.0−19.2
ArXivMath 06/2026Math57.676.7+19.1
DeepsecBench-AA (Artificial Analysis)Cybersecurity31.212.5−18.7
Creative Writing v3 EloChat & Writing15401907+18.6
Mystery Game PuzzlesGames43.025.0−18.0
Harvey's Legal Agent BenchmarkLegal6.723.8+17.1
RuneBenchGames5.74.6−16.8
KernelBench CUDA Normalized (RTX PRO 6000)Code47.030.3−16.7
SAFE-Teen 106 worst branch: Relationships (under-18 prompt)Safety19.83.2+16.6
IOI (Vals v2)Code40.356.6+16.3
BuseyBench SVGMultimodal4.15.7+16.2
ChartographyMultimodal11.627.6+16.0
SAFE-Teen 106 worst branch: Urgent Safety (under-18 prompt)Safety27.211.3+15.8
SheetBench-50Finance & Accounting58.442.8−15.6
AA-Omniscience IndexIndexes-5.325.0+15.2
SimpleBenchGeneral QA66.781.8+15.1
ArXivMath 05/2026Math58.373.3+15.0
GeneBench-Pro Public Release SubsetScience11.026.0+15.0
MysteryMechanismScience21.236.0+14.9
Humanity's Last Exam DiamondGeneral QA19.133.7+14.6
τ³-Banking (Mercor run, bm25_grep retrieval)Agents37.852.2+14.4
Legal Research BenchLegal41.355.3+13.9
BrokenArXiv 06/2026Math39.453.2+13.9
ITBench-AAAgents46.933.2−13.7
CharXiv Extended (Mercor held-out)Multimodal68.981.8+12.9
APEX-SWECode51.939.1−12.8
MedXpertQA MM (200-question set)Health & Medicine63.576.3+12.8
LLM Creative Story WritingChat & Writing-0.50.6+11.8
CritPt (avg@5)Science14.326.0+11.7
MedXpertQA MM Extended (Mercor held-out)Health & Medicine48.460.1+11.7
Design Creativity BenchCode43.647.2+11.4
AutomationBench-AA 1.0.6Agents68.957.9−11.0
Chess Puzzles (task version 2.0)Games48.037.0−11.0
Eval Connections One-Shot with TrapsGames73.084.0+11.0
APEX-Agents 1.1Professional Work50.561.3+10.8
Conceptual Reasoning IndexIndexes58.466.2+10.7
Excel Modeling BenchmarkFinance & Accounting57.267.4+10.2
Bug Hunt BenchCode20.630.7+10.0
Terminal-Bench Science 0.1 (Vals)Science0.010.0+10.0
Tax Agent BenchFinance & Accounting62.572.4+10.0
dfbench Detection RecallCybersecurity57.347.4−9.9
AA-Briefcase EloAgents14241589+9.6
AA-PersonalityBench Expressiveness indexChat & Writing61.971.4+9.6
DAXBench v2Code75.585.1+9.6
Longform Creative WritingChat & Writing74.082.8+8.8
MMMU-Pro Extended (Mercor held-out)Multimodal25.033.8+8.8
Artificial Analysis Intelligence Index v4.3.2Indexes39.548.1+8.6
Artificial Analysis Intelligence Index v4.3Indexes39.548.2+8.6
BrokenArXiv 05/2026Math28.536.5+8.0
CadQueryEval (Corrected Geometry)Code76.084.0+8.0
Creative Writing v3 Rubric ScoreChat & Writing14.115.6+8.0
SciCode 1.0.1 (Artificial Analysis)Code51.959.7+7.9
GDP.pdfProfessional Work19.827.6+7.8
APEX-1 (Mercor, mean score)Professional Work65.573.0+7.5
CWE-Bench-AA (Artificial Analysis)Cybersecurity68.360.8−7.5
Vision Arena OverallMultimodal12631291+7.3
SAFE-Teen 106: Relationships (no prompt)Safety27.134.4−7.2
Harvey LAB-AA v1.1 (hallucination-gated)Legal1.78.9+7.2
FrontierMath Tiers 1-3 (v2) - PrivateMath67.474.4+7.0
ACCoRDGeneral QA68.175.1+7.0
LMCAGeneral QA47.053.9+6.9
Vals Index v2Indexes57.964.5+6.7
DTBenchGeneral QA89.996.5+6.7
Terminal-Bench 4.0 (Artificial Analysis)Code26.833.3+6.6
Finance Agent v2Finance & Accounting53.560.0+6.5
Humanity's Last ExamGeneral QA36.843.2+6.4
Arena Image-to-WebDevCode16101637+5.9
SAFE-Teen 106 worst branch: Urgent Safety (no prompt)Safety31.425.6+5.8
Decision-Bench v1.1Professional Work25.319.8−5.4
APEX-AccountingFinance & Accounting48.153.5+5.4
ALE-BenchCode18291981+5.4
BizAgentBench (strict success)Agents43.338.0−5.3
SAFE-Teen 106: Urgent Safety (under-18 prompt)Safety7.92.7+5.2
DistressBenchSafety79.884.9+5.1
SAFE-Teen 106 worst branch: Relationships (no prompt)Safety51.156.2−5.1
GBENCH One-shot Coding GScoreGames48.052.7+4.7
MineBench — Bradley-Terry RatingGames18221747−4.5
LMArena Code Arena WebDevCode16211655+4.5
Extended NYT Connections 940 (Quadratic v1)Games89.685.1−4.5
Legacy-Bench (Fireworks, avg@3)Code57.762.0+4.3
RiemannBenchMath24.028.0+4.0
ProofBench v1.1Math54.058.0+4.0
AA-LCR Extended (Mercor held-out)Search70.866.9−3.9
CharXiv (descriptive and reasoning combined)Multimodal90.193.9+3.8
LMArena Text Arena EloChat & Writing14751496+3.7
SAFE-Teen 106: Relationships (under-18 prompt)Safety4.20.6+3.6
GDPval-AA v2.1Professional Work16001681+3.6
Artificial Analysis Cyber Index v1Indexes40.844.3+3.5
Terminal-Bench 4.0Code31.227.8−3.4
FrontierCode 1.1 Main (Artificial Analysis)Code44.040.6−3.4
SWE-bench Multimodal (Vals v1, 480 Tasks)Code46.543.1−3.3
LiveBench 2.0Agents78.074.8−3.3
Westworld Finance Diligence BenchFinance & Accounting48.345.2−3.1
SRE Bench — Fully SolvedCode0.83.8+3.1
Code2Video-AA EloCode10531019−2.8
AA-AnalystAgent (pass^5)Professional Work35.037.5+2.5
LMArena Agent Arena SteerabilityAgents-3.31.6+2.4
LMArena Agent Arena Code Net Improvement (Steering Burden)Agents2.73.8+2.3
LMArena Agent Arena Chat Net Improvement (Steering Burden)Agents1.62.6+2.3
SAFE-Teen 106: Urgent Safety (no prompt)Safety11.39.0+2.3
Arena Alignment IndexSafety74.877.1+2.3
LMArena Agent Arena Chat Bash ReliabilityAgents-1.03.4+2.2
ComplexConstraintsChat & Writing49.751.9+2.2
LMArena Agent Arena Code Steering BurdenAgents2.36.7+2.2
LMArena Agent Arena Work Steering BurdenAgents1.55.7+2.1
ProgramBenchCode0.52.5+2.0
TaxCalcBench TY25Finance & Accounting4.06.0+2.0
LMArena Agent Arena Work Net Improvement (Steering Burden)Agents2.53.3+1.9
Terminal-Bench-Science 0.1Science9.011.0+1.9
Financial Coaching BenchFinance & Accounting72.871.0−1.8
Terminal-Bench 2.1Code90.688.8−1.8
Code MigrationCode45.647.4+1.8
ArXivMath 08/2026 (Native Harnesses)Math43.945.6+1.8
AIME 2025Math98.3100.0+1.7
Ramp Accounting BenchFinance & Accounting40.842.5+1.7
LMArena Agent Arena Code Bash ReliabilityAgents-0.23.1+1.6
GDPval-AA v2Professional Work16321662+1.5
Artificial Analysis Progress Index v1Indexes320324+1.5
LLM Debate BenchmarkChat & Writing15821593+1.5
SpeciEval (Refusal-Aware Retries, Historical)Safety97.498.8+1.4
SpeciEval (Pooled English Runs)Safety97.498.7+1.4
Epoch Capabilities Index (ECI)Indexes156158+1.2
DeepSWE v1.1Code74.275.4+1.2
LMArena Agent Arena Chat Steering BurdenAgents-0.51.8+1.1
Vibe Code Bench v1.1Code84.785.9+1.1
LMArena Agent Arena Work Bash ReliabilityAgents0.12.0+1.0
CyberBench v1.1 OverallCybersecurity73.772.7−1.0
LMArena Agent Arena Chat Confirmed SuccessAgents4.93.1−0.9
LMArena Agent Arena Bash RecoveryAgents7.96.1−0.9
OTIS Mock AIME 2024-2025Math98.399.2+0.8
WeirdML v3Science5.86.7+0.8
LMArena Agent Arena Chat Praise vs ComplaintAgents3.14.6+0.8
Access to Justice BenchmarkLegal72.972.2−0.7
Immigration BenchLegal65.665.0−0.6
LMArena Agent Arena Code Confirmed SuccessAgents6.65.5−0.5
LMArena Agent Arena Work Praise vs ComplaintAgents3.72.7−0.5
LMArena Agent Arena Work Confirmed SuccessAgents5.75.0−0.3
LMArena Agent Arena Code Praise vs ComplaintAgents3.02.6−0.2
LMArena Agent Arena Net ImprovementAgents3.93.8−0.1
GameBench 2Games68.468.5+0.1
LMArena Agent Arena Chat Tool HallucinationAgents0.30.5+0.1
LMArena Agent Arena Code Tool HallucinationAgents0.50.7+0.1
LMArena Agent Arena Work Tool HallucinationAgents0.50.7+0.1
CWE-bench v1 Programmatic Pass@1Cybersecurity55.055.00.0

DeepSeek V4.1 FlashvsMuse Spark 1.3normalised score marginsData Nov 15, 2026Data: Vals AI, Artificial Analysis, Arena and 37 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…