Atlas

Compare Models

Claude Opus 5 vs Gemini 4 Neon

Gemini 4 Neon leads by 5.0 capability points (≈95% interval +2.3 to +7.8): a clear lead over Claude Opus 5 · P(higher) ≈ 100%

Wins 46 of 48 shared benchmarks (2 lost)

Capability vs. relative cost
Claude Opus 5Gemini 4 NeonClosed weightsOpen weights

Benchmark wins2 – 46 · 48 shared

Comparing matching setups only (4 benchmarks excluded) · suites counted once
← Claude Opus 5Gemini 4 Neon →
Data: Artificial Analysis, Arena, Vals AI and 11 othersatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Gemini 4 Neon minus Claude Opus 5.

Humanity's Sixth SensePerception & media › Spatial & visual reasoning30.575.9+45.4
Terminal-Bench-Science 0.1Science › Multidisciplinary28.670.5+41.9
Chess Puzzles (task version 2.0)Games43.083.0+40.0
Terminal-Bench Science 0.1 (Vals)Science › Multidisciplinary27.162.9+35.7
Bug Hunt BenchSoftware › Software engineering25.760.0+34.3
Vending-Bench 2Business simulations1118216843+33.6
AutomationBench 1.0.6 (Zapier)Tool & computer use › Customer service & operations26.956.3+29.4
Furniture Assembly BenchmarkPerception & media › Spatial & visual reasoning60.889.2+28.3
DAYJOB: Finance (October 2026 rescoring)Professional › Finance, accounting & tax23.850.5+26.7
Humanity's Last Exam DiamondKnowledge & reasoning › General knowledge44.871.3+26.5
ExploitGym - release-reported percentage, cohort unspecifiedSoftware › Cybersecurity22.148.3+26.2
Mystery Game PuzzlesGames59.085.0+26.0
DAYJOB: Healthcare (October 2026 rescoring)Health & medicine22.047.2+25.2
FrontierMath Tier 4 (v2) - PrivateMath73.297.6+24.4
LMArena Agent Arena Net Improvement (Steering Burden)Tool & computer use › General agents6.318.4+23.7
AutomationBench-AA 1.0.6Tool & computer use › Customer service & operations56.679.3+22.8
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & medicine13.931.7+17.8
Artificial Analysis Intelligence Index v5.0.1Composite indexes44.261.4+17.2
Terminal-Bench 4.0 (Artificial Analysis)Software › Software engineering46.563.6+17.2
LMArena Agent Arena Praise vs ComplaintTool & computer use › General agents10.344.5+17.1
τ³-Banking (Mercor run, bm25_grep retrieval)Tool & computer use › Customer service & operations44.059.5+15.5
Vals Science IndexScience › Multidisciplinary47.962.6+14.7
SimpleBenchKnowledge & reasoning › General knowledge80.694.4+13.8
FrontierCode 2.0 (Artificial Analysis)Software › Software engineering31.244.4+13.3
APEX-Agents 1.1Professional › General knowledge work77.088.9+11.9
GDP.pdf (Artificial Analysis, All-pass)Professional › General knowledge work21.633.4+11.8
SWE-sweepSoftware › Software engineering1.312.8+11.5
LMArena Code Arena WebDevSoftware › Frontend & design16911764+9.6
LMArena Agent Arena Steering BurdenTool & computer use › General agents2.421.2+9.4
EnterpriseBench: CoreCraftTool & computer use › Customer service & operations68.777.4+8.7
AA-Omniscience v1.1 IndexKnowledge & reasoning › General knowledge37.654.8+8.6
GPQA DiamondScience › Multidisciplinary87.996.3+8.4
MysteryMechanismMath37.445.0+7.7
LMArena Agent Arena Bash ReliabilityTool & computer use › General agents8.7-6.0−7.4
CritPt v2 (agentic, avg@5)Science › Physical sciences71.478.0+6.6
APEX-Accounting 1.1Professional › Finance, accounting & tax64.370.8+6.5
FrontierMath Tiers 1-3 (v2) - PrivateMath85.691.2+5.6
Vibe Code Bench v1.1Software › Frontend & design88.493.7+5.3
LMArena Text Arena EloLanguage & writing › General assistance14901518+5.0
LMArena Agent Arena Confirmed SuccessTool & computer use › General agents10.119.2+4.6
AA-Briefcase v1.1 EloProfessional › General knowledge work16291699+3.8
Tax Agent BenchProfessional › Finance, accounting & tax76.779.8+3.1
Chartography (November 2026 rescoring)Perception & media › Documents & charts84.486.6+2.2
Vals RSI Index v1.1 (four tasks)Software › Data & ML33.034.4+1.4
OTIS Mock AIME 2024-2025Math98.9100.0+1.1
Legal Research BenchProfessional › Legal55.356.3+1.0
BioMysteryBench (Vals Terminus 2)Science › Life sciences79.380.0+0.7
LMArena Agent Arena Tool HallucinationTool & computer use › Tool calling1.01.00.0
Data: Artificial Analysis, Arena, Vals AI and 11 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…