Atlas

Compare Models

Claude Sonnet 5.5 vs Gemini 4 Neon

Gemini 4 Neon leads by 2.4 capability points (≈95% interval −0.5 to +5.3): statistically tied · P(higher) ≈ 95%

Wins 53 of 62 shared benchmarks (8 lost, 1 tied)

Capability vs. relative cost
Claude Sonnet 5.5Gemini 4 NeonClosed weightsOpen weights

Benchmark wins8 – 53 · 62 shared, 1 tied

Comparing matching setups only (4 benchmarks excluded) · suites counted once
← Claude Sonnet 5.5Gemini 4 Neon →
Data: Artificial Analysis, Surge AI, Arena and 12 othersatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Gemini 4 Neon minus Claude Sonnet 5.5.

Vending-Bench 2Business simulations1116516843+33.6
SimpleQA Verified (Epoch anti-abstention)Knowledge & reasoning › General knowledge46.575.2+28.7
Chess Puzzles (task version 2.0)Games59.083.0+24.0
Mystery Game PuzzlesGames65.085.0+20.0
SimpleBenchKnowledge & reasoning › General knowledge75.994.4+18.5
SWE-Marathon v1.2Software › Software engineering80.662.5−18.1
Humanity's Last Exam DiamondKnowledge & reasoning › General knowledge53.371.3+18.0
Terminal-Bench-Science 0.1Science › Multidisciplinary52.470.0+17.6
FrontierMath Tier 4 (v2) - PrivateMath80.597.6+17.1
Terminal-Bench Science 0.1 (Vals)Science › Multidisciplinary47.162.9+15.7
Agents' Last Exam v1.1 ScoreProfessional › General knowledge work60.475.8+15.4
DAYJOB: AccountingProfessional › Finance, accounting & tax35.249.6+14.4
Furniture Assembly BenchmarkPerception & media › Spatial & visual reasoning75.089.2+14.2
GDP.xlsxProfessional › General knowledge work29.143.1+14.0
DAYJOB: Healthcare (October 2026 rescoring)Health & medicine34.047.2+13.2
sudo L7 (October 2026 rescoring)Software › Software engineering50.638.3−12.3
AA-Omniscience v1.1 IndexKnowledge & reasoning › General knowledge30.554.8+12.1
ChartographyPerception & media › Documents & charts61.673.5+11.9
Artificial Analysis Legal IndexProfessional › Legal52.263.8+11.6
Bug Hunt BenchSoftware › Software engineering48.960.0+11.1
CursorBench 4.1Software › Software engineering54.465.4+11.0
LiveBench 2.0 (2026-11-06 release)Composite indexes76.887.4+10.6
DAYJOB: Finance (October 2026 rescoring)Professional › Finance, accounting & tax40.550.5+10.0
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & medicine41.731.7−10.0
CWE-bench v2 Pass@1Software › Cybersecurity61.470.7+9.3
Artificial Analysis Finance & Accounting IndexProfessional › Finance, accounting & tax55.865.1+9.3
HLAB-Vals-1.1Professional › Legal10.419.6+9.2
LMArena Agent Arena Bash ReliabilityTool & computer use › General agents11.9-6.0−9.0
FrontierCode 2.0 (Artificial Analysis)Software › Software engineering35.644.4+8.8
Artificial Analysis Engineering IndexScience › Engineering58.566.8+8.3
Artificial Analysis Economics IndexComposite indexes54.161.9+7.9
LMArena Text Arena EloLanguage & writing › General assistance14741518+7.8
GDP.pdf (Artificial Analysis, All-pass)Professional › General knowledge work25.833.4+7.6
AutomationBench-AA 1.0.6Tool & computer use › Customer service & operations71.879.3+7.6
RiemannBench (November 2026 rescoring)Math76.884.0+7.2
LMArena Agent Arena Praise vs ComplaintTool & computer use › General agents30.744.5+6.9
Artificial Analysis Intelligence Index v5.0.1Composite indexes54.561.4+6.9
EnterpriseBench: CoreCraftTool & computer use › Customer service & operations70.877.4+6.6
Terminal-Bench 4.0 (Artificial Analysis)Software › Software engineering57.163.6+6.6
APEX-Accounting 1.1Professional › Finance, accounting & tax64.570.8+6.3
Artificial Analysis Healthcare & Medical IndexHealth & medicine54.260.1+5.9
Artificial Analysis Strategy & Ops IndexProfessional › General knowledge work58.163.8+5.7
ComplexConstraints (November 2026 rescoring)Language & writing › Instruction following67.272.9+5.7
LMArena Agent Arena Net Improvement (Steering Burden)Tool & computer use › General agents15.618.4+5.6
τ³-Banking (Mercor run, bm25_grep retrieval)Tool & computer use › Customer service & operations54.059.5+5.5
APEX-Agents 1.1Professional › General knowledge work83.788.9+5.2
HANDBOOK.md Agents v1.1Tool & computer use › Customer service & operations47.752.3+4.6
MKA (Artificial Analysis)Professional › General knowledge work52.056.5+4.5
MysteryMechanismMath49.145.0−4.1
LMArena Agent Arena Steering BurdenTool & computer use › General agents14.821.2+3.2
FrontierMath Tiers 1-3 (v2) - PrivateMath88.891.2+2.5
CritPt v2 (agentic, avg@5)Science › Physical sciences75.177.4+2.3
GDP.pdf (November 2026 rescoring)Professional › General knowledge work51.253.4+2.2
AA-Briefcase v1.1 EloProfessional › General knowledge work17391699−2.2
DAYJOB: TaxProfessional › Finance, accounting & tax10.812.8+2.0
Vibe Code Bench v1.1Software › Frontend & design92.493.7+1.3
LMArena Code Arena WebDevSoftware › Frontend & design17731764−1.2
LMArena Agent Arena Confirmed SuccessTool & computer use › General agents17.619.2+0.8
DAYJOB: Law (November 2026 release)Professional › Legal26.826.2−0.6
Excel Modeling Benchmark v1.1Professional › Finance, accounting & tax75.976.5+0.6
LMArena Agent Arena Tool HallucinationTool & computer use › Tool calling1.01.00.0
OTIS Mock AIME 2024-2025Math100.0100.00.0
Data: Artificial Analysis, Surge AI, Arena and 12 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…