Atlas

Compare Models

Claude Haiku 5.5 vs Gemini 4 Neon

Gemini 4 Neon leads by 7.7 capability points (≈95% interval +5.1 to +10.4): a clear lead over Claude Haiku 5.5 · P(higher) ≈ 100%

Wins 57 of 59 shared benchmarks (2 lost)

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

Benchmark wins2 – 57 · 59 shared

Comparing matching setups only (3 benchmarks excluded) · suites counted once
← Claude Haiku 5.5Gemini 4 Neon →
Data: Artificial Analysis, Vals AI, Arena 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 Claude Haiku 5.5.

Vending-Bench 2Business simulations360216843+78.5
Mystery Game PuzzlesGames30.085.0+55.0
Terminal-Bench Science 0.1 (Vals)Science › Multidisciplinary10.062.9+52.9
Terminal-Bench-Science 0.1Science › Multidisciplinary18.170.0+51.9
SimpleQA Verified (Epoch anti-abstention)Knowledge & reasoning › General knowledge23.875.2+51.4
FrontierMath Tier 4 (v2) - PrivateMath46.397.6+51.2
AutomationBench 1.0.6 (Zapier)Tool & computer use › Customer service & operations12.056.3+44.3
AutomationBench-AA 1.0.6Tool & computer use › Customer service & operations36.079.3+43.4
Humanity's Last Exam DiamondKnowledge & reasoning › General knowledge29.671.3+41.7
Furniture Assembly BenchmarkPerception & media › Spatial & visual reasoning47.589.2+41.7
Bug Hunt BenchSoftware › Software engineering20.560.0+39.5
LMArena Agent Arena Net Improvement (Steering Burden)Tool & computer use › General agents-0.618.4+37.1
SimpleBenchKnowledge & reasoning › General knowledge59.294.4+35.2
Terminal-Bench 4.0 (Artificial Analysis)Software › Software engineering29.363.6+34.3
Chess Puzzles (task version 2.0)Games55.083.0+28.0
Code Migration v1.1 (Vals Index v2.2 subset)Software › Software engineering50.275.8+25.6
Terminal-Bench 4.0Software › Software engineering36.461.6+25.3
Artificial Analysis Legal IndexProfessional › Legal38.763.8+25.1
LMArena Agent Arena Praise vs ComplaintTool & computer use › General agents-5.544.5+25.0
Artificial Analysis Intelligence Index v5.0.1Composite indexes36.561.4+24.9
Artificial Analysis Strategy & Ops IndexProfessional › General knowledge work39.163.8+24.7
Artificial Analysis Engineering IndexScience › Engineering42.566.8+24.3
Vals Science IndexScience › Multidisciplinary38.562.6+24.1
Artificial Analysis Healthcare & Medical IndexHealth & medicine37.560.1+22.6
LMArena Code Arena WebDevSoftware › Frontend & design15961764+22.4
Harvey LAB (Mercor run, undisclosed ~213-task cohort)Professional › Legal4.926.6+21.7
Artificial Analysis Finance & Accounting IndexProfessional › Finance, accounting & tax43.665.1+21.5
AA-Omniscience v1.1 IndexKnowledge & reasoning › General knowledge12.354.8+21.2
APEX-Agents 1.1Professional › General knowledge work68.388.9+20.6
Artificial Analysis Economics IndexComposite indexes42.761.9+19.2
LiveBench 2.0 (2026-11-06 release)Composite indexes68.987.4+18.5
Tax Agent BenchProfessional › Finance, accounting & tax62.279.8+17.6
CursorBench 4.1Software › Software engineering48.465.4+17.0
LMArena Text Arena EloLanguage & writing › General assistance14211518+16.9
τ³-Banking (Mercor run, bm25_grep retrieval)Tool & computer use › Customer service & operations43.059.5+16.5
FrontierMath Tiers 1-3 (v2) - PrivateMath75.191.2+16.1
MKA (Artificial Analysis)Professional › General knowledge work41.056.5+15.5
HLAB-Vals-1.1Professional › Legal4.219.6+15.4
FrontierCode 2.0 (Artificial Analysis)Software › Software engineering29.944.4+14.5
SWE-Marathon v1.2Software › Software engineering48.162.5+14.4
GDP.pdf (Artificial Analysis, All-pass)Professional › General knowledge work20.833.4+12.6
Legal Research BenchProfessional › Legal43.856.3+12.5
Vals Index v2.2Composite indexes58.770.9+12.1
CritPt v2 (agentic, avg@5)Science › Physical sciences66.377.4+11.1
Excel Modeling Benchmark v1.1Professional › Finance, accounting & tax65.576.5+11.0
Vals Finance IndexProfessional › Finance, accounting & tax57.468.3+10.9
LMArena Agent Arena Steering BurdenTool & computer use › General agents0.821.2+10.2
MysteryMechanismMath35.145.0+9.9
AA-Briefcase v1.1 EloProfessional › General knowledge work15231699+9.6
BioMysteryBench (Vals Terminus 2)Science › Life sciences70.480.0+9.6
LMArena Agent Arena Confirmed SuccessTool & computer use › General agents0.219.2+9.5
APEX-Accounting 1.1Professional › Finance, accounting & tax62.270.8+8.6
Vals Legal IndexProfessional › Legal70.378.4+8.1
LMArena Agent Arena Bash ReliabilityTool & computer use › General agents6.6-6.0−6.3
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & medicine35.031.7−3.3
Vibe Code Bench v1.1Software › Frontend & design90.493.7+3.3
Finance Agent v2.1Professional › Finance, accounting & tax65.468.6+3.2
LMArena Agent Arena Tool HallucinationTool & computer use › Tool calling-1.51.0+1.3
OTIS Mock AIME 2024-2025Math98.9100.0+1.1
Data: Artificial Analysis, Vals AI, Arena 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…