Atlas

Compare Models

DeepSeek V4.1 Flash 1017 vs Gemini 4 Hydrogen

Gemini 4 Hydrogen leads by 4.9 capability points (≈95% interval +2.3 to +7.6): a clear lead over DeepSeek V4.1 Flash 1017 · P(higher) ≈ 100%

Wins 60 of 63 shared benchmarks (3 lost)

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

Benchmark wins3 – 60 · 63 shared

Comparing matching setups only (5 benchmarks excluded) · suites counted once
← DeepSeek V4.1 Flash 1017Gemini 4 Hydrogen →
Data: Artificial Analysis, Vals AI, Epoch AI and 14 othersatlas.village.dev

Margins

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

FrontierMath Tier 4 (v2) - PrivateMath46.387.8+41.5
Humanity's Last Exam DiamondKnowledge & reasoning › General knowledge20.858.8+38.0
Humanity's Sixth SensePerception & media › Spatial & visual reasoning30.163.4+33.3
Terminal-Bench-Science 0.1Science › Multidisciplinary10.541.4+31.0
SimpleQA Verified (Epoch anti-abstention)Knowledge & reasoning › General knowledge42.570.9+28.4
DAYJOB: AccountingProfessional › Finance, accounting & tax7.633.6+26.0
DAYJOB: Finance (October 2026 rescoring)Professional › Finance, accounting & tax12.538.5+26.0
CritPt v2 (agentic, avg@5)Science › Physical sciences52.378.0+25.7
Terminal-Bench 4.0 (Artificial Analysis)Software › Software engineering28.854.0+25.3
BuseyBench SVGPerception & media › Generation4.36.9+25.1
LMArena Agent Arena Net Improvement (Steering Burden)Tool & computer use › General agents0.813.5+24.8
Artificial Analysis Engineering IndexScience › Engineering37.660.1+22.5
GDP.xlsxProfessional › General knowledge work18.440.9+22.5
AA-Omniscience v1.1 IndexKnowledge & reasoning › General knowledge1.546.3+22.4
Terminal-Bench 4.0Software › Software engineering31.854.0+22.2
HANDBOOK.md Agents v1.1Tool & computer use › Customer service & operations24.044.6+20.6
FrontierCode 2.0 - Pass RateSoftware › Software engineering23.043.2+20.2
Artificial Analysis Legal IndexProfessional › Legal37.756.7+19.0
FrontierCode 2.0 - ScoreSoftware › Software engineering20.238.9+18.7
DAYJOB: Healthcare (October 2026 rescoring)Health & medicine6.024.0+18.0
FrontierMath Tiers 1-3 (v2) - PrivateMath71.288.1+16.8
Artificial Analysis Economics IndexComposite indexes38.555.0+16.6
Mystery Game PuzzlesGames49.065.0+16.0
Artificial Analysis Finance & Accounting IndexProfessional › Finance, accounting & tax42.258.1+15.9
GDP.pdf (Artificial Analysis, All-pass)Professional › General knowledge work17.232.8+15.6
Artificial Analysis Coding Agent Index v2.1Software › Software engineering36.852.3+15.5
LMArena Agent Arena Praise vs ComplaintTool & computer use › General agents-1.529.3+15.4
ARC-AGI-2Knowledge & reasoning › Abstract reasoning78.192.5+14.4
Code Migration v1.1 (Vals Index v2.2 subset)Software › Software engineering68.582.7+14.2
Chess Puzzles (task version 2.0)Games46.060.0+14.0
Artificial Analysis Healthcare & Medical IndexHealth & medicine41.353.3+12.0
GBENCH One-shot Coding GScoreGames55.466.8+11.4
Ramp SWE-Bench v2Software › Software engineering60.871.6+10.8
CWE-bench v2 Pass@1Software › Cybersecurity57.668.1+10.5
Bug Hunt BenchSoftware › Software engineering40.731.1−9.6
Legal Research BenchProfessional › Legal45.253.8+8.7
Artificial Analysis Intelligence Index v5.0.1Composite indexes32.641.2+8.6
Artificial Analysis Strategy & Ops IndexProfessional › General knowledge work49.257.7+8.5
LMArena Agent Arena Steering BurdenTool & computer use › General agents2.317.4+7.6
Agents' Last Exam v1.1 (ALE v1.1)Professional › General knowledge work36.143.4+7.4
Vals Index v2.2Composite indexes61.368.5+7.2
DAYJOB: Law (November 2026 release)Professional › Legal2.89.7+6.9
LiveBench 2.0 (2026-11-06 release)Composite indexes70.377.0+6.7
LMArena Agent Arena Confirmed SuccessTool & computer use › General agents2.215.7+6.7
Excel Modeling Benchmark v1.1Professional › Finance, accounting & tax69.275.8+6.6
CursorBench 4.0Software › Software engineering47.754.3+6.6
AutomationBench-AA 1.0.6Tool & computer use › Customer service & operations68.574.4+5.9
Epoch Capabilities Index (ECI)Composite indexes159167+5.9
LMArena Code Arena WebDevSoftware › Frontend & design16591703+5.9
Vals Finance IndexProfessional › Finance, accounting & tax60.966.1+5.2
AA-Briefcase v1.1 EloProfessional › General knowledge work14161497+4.4
ARC-AGI-1 Semi-PrivateKnowledge & reasoning › Abstract reasoning94.598.5+4.0
Tax Agent BenchProfessional › Finance, accounting & tax74.978.4+3.5
HLAB-Vals-1.1Professional › Legal14.211.3−2.9
LMArena Agent Arena Bash ReliabilityTool & computer use › General agents1.1-4.5−2.8
DAYJOB: TaxProfessional › Finance, accounting & tax1.23.6+2.4
Finance Agent v2.1Professional › Finance, accounting & tax62.965.1+2.2
Vals Legal IndexProfessional › Legal71.973.6+1.8
DualEntry Accounting AI BenchmarkProfessional › Finance, accounting & tax82.283.2+1.0
OTIS Mock AIME 2024-2025Math99.4100.0+0.6
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & medicine26.727.2+0.6
LMArena Agent Arena Tool HallucinationTool & computer use › Tool calling-0.10.7+0.4
Vibe Code Bench v1.1Software › Frontend & design87.787.80.0
Data: Artificial Analysis, Vals AI, Epoch AI and 14 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…