Atlas

Compare Models

DeepSeek V4.1 Flash 1017 vs Gemini 4 Neon

Gemini 4 Neon leads by 8.2 capability points (≈95% interval +5.3 to +11.1): a clear lead over DeepSeek V4.1 Flash 1017 · P(higher) ≈ 100%

Wins 60 of 61 shared benchmarks (1 lost)

Capability vs. relative cost
DeepSeek V4.1 Flash 1017Gemini 4 NeonClosed weightsOpen weights

Benchmark wins1 – 60 · 61 shared

Comparing matching setups only (2 benchmarks excluded) · suites counted once
← DeepSeek V4.1 Flash 1017Gemini 4 Neon →
Data: Artificial Analysis, Vals AI, Surge AI and 9 othersatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Gemini 4 Neon minus DeepSeek V4.1 Flash 1017.

Terminal-Bench-Science 0.1Science › Multidisciplinary10.565.7+55.2
FrontierMath Tier 4 (v2) - PrivateMath46.397.6+51.2
Humanity's Last Exam DiamondKnowledge & reasoning › General knowledge20.871.3+50.5
Humanity's Sixth SensePerception & media › Spatial & visual reasoning30.175.9+45.8
RiemannBench (November 2026 rescoring)Math38.484.0+45.6
DAYJOB: AccountingProfessional › Finance, accounting & tax7.649.6+42.0
DAYJOB: Healthcare (October 2026 rescoring)Health & medicine6.047.2+41.2
DAYJOB: Finance (October 2026 rescoring)Professional › Finance, accounting & tax12.550.5+38.0
Chess Puzzles (task version 2.0)Games46.083.0+37.0
Mystery Game PuzzlesGames49.085.0+36.0
LMArena Agent Arena Net Improvement (Steering Burden)Tool & computer use › General agents0.818.4+34.4
Terminal-Bench 4.0 (Artificial Analysis)Software › Software engineering28.861.6+32.8
SimpleQA Verified (Epoch anti-abstention)Knowledge & reasoning › General knowledge42.575.2+32.7
Terminal-Bench 4.0Software › Software engineering31.861.6+29.8
Artificial Analysis Engineering IndexScience › Engineering37.666.8+29.2
Artificial Analysis Intelligence Index v5.0.1Composite indexes32.661.4+28.8
HANDBOOK.md Agents v1.1Tool & computer use › Customer service & operations24.052.3+28.3
AA-Omniscience v1.1 IndexKnowledge & reasoning › General knowledge1.554.8+26.6
FrontierCode 2.0 (Artificial Analysis)Software › Software engineering16.843.3+26.4
Artificial Analysis Legal IndexProfessional › Legal37.763.8+26.0
GDP.xlsxProfessional › General knowledge work18.443.1+24.7
CritPt v2 (agentic, avg@5)Science › Physical sciences52.376.6+24.3
Artificial Analysis Economics IndexComposite indexes38.561.9+23.4
DAYJOB: Law (November 2026 release)Professional › Legal2.826.2+23.4
LMArena Agent Arena Praise vs ComplaintTool & computer use › General agents-1.544.5+23.0
Artificial Analysis Finance & Accounting IndexProfessional › Finance, accounting & tax42.265.1+22.9
Artificial Analysis Coding Agent Index v2.1Software › Software engineering36.858.6+21.8
FrontierMath Tiers 1-3 (v2) - PrivateMath71.291.2+20.0
Bug Hunt BenchSoftware › Software engineering40.760.0+19.3
EnterpriseBench: CoreCraftTool & computer use › Customer service & operations58.577.4+18.9
Artificial Analysis Healthcare & Medical IndexHealth & medicine41.360.1+18.8
MKA (Artificial Analysis)Professional › General knowledge work38.556.5+17.9
τ³-Banking (Mercor run, bm25_grep retrieval)Tool & computer use › Customer service & operations41.659.5+17.9
LiveBench 2.0 (2026-11-06 release)Composite indexes70.387.4+17.2
GDP.pdf (Artificial Analysis, All-pass)Professional › General knowledge work17.233.4+16.2
Artificial Analysis Strategy & Ops IndexProfessional › General knowledge work49.263.8+14.6
LMArena Code Arena WebDevSoftware › Frontend & design16591764+14.0
CWE-bench v2 Pass@1Software › Cybersecurity57.670.7+13.1
AA-Briefcase v1.1 EloProfessional › General knowledge work14161632+11.8
DAYJOB: TaxProfessional › Finance, accounting & tax1.212.8+11.6
APEX-Accounting 1.1Professional › Finance, accounting & tax59.370.8+11.5
Legal Research BenchProfessional › Legal45.256.3+11.1
AutomationBench-AA 1.0.6Tool & computer use › Customer service & operations68.579.3+10.8
Vals Index v2.2Composite indexes61.370.9+9.6
APEX-Agents 1.1Professional › General knowledge work79.488.9+9.5
LMArena Agent Arena Steering BurdenTool & computer use › General agents2.321.2+9.4
Agents' Last Exam v1.1 ScoreProfessional › General knowledge work66.675.8+9.2
LMArena Agent Arena Confirmed SuccessTool & computer use › General agents2.219.2+8.5
Harvey LAB (Mercor run, undisclosed ~213-task cohort)Professional › Legal18.826.6+7.8
Vals Finance IndexProfessional › Finance, accounting & tax60.968.3+7.4
Excel Modeling Benchmark v1.1Professional › Finance, accounting & tax69.276.5+7.3
Code Migration v1.1 (Vals Index v2.2 subset)Software › Software engineering68.575.8+7.3
Vals Legal IndexProfessional › Legal71.978.4+6.5
Vibe Code Bench v1.1Software › Frontend & design87.793.7+6.0
Finance Agent v2.1Professional › Finance, accounting & tax62.968.6+5.6
HLAB-Vals-1.1Professional › Legal14.219.6+5.4
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & medicine26.731.7+5.0
Tax Agent BenchProfessional › Finance, accounting & tax74.979.8+4.9
LMArena Agent Arena Bash ReliabilityTool & computer use › General agents1.1-6.0−3.5
OTIS Mock AIME 2024-2025Math99.4100.0+0.6
LMArena Agent Arena Tool HallucinationTool & computer use › Tool calling-0.11.0+0.5
Data: Artificial Analysis, Vals AI, Surge AI and 9 othersatlas.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…