Atlas

Compare Models

Claude Haiku 5.5 vs GPT-6.2 Sol

GPT-6.2 Sol leads by 5.7 capability points (≈95% interval +3.3 to +8.2): a clear lead over Claude Haiku 5.5 · P(higher) ≈ 100%

Wins 51 of 63 shared benchmarks (11 lost, 1 tied)

Capability vs. relative cost
Claude Haiku 5.5GPT-6.2 SolClosed weightsOpen weights

Benchmark wins11 – 51 · 63 shared, 1 tied

Comparing matching setups only (10 benchmarks excluded) · suites counted once
← Claude Haiku 5.5GPT-6.2 Sol →
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 GPT-6.2 Sol minus Claude Haiku 5.5.

ARC-AGI-3 (Semi-Private)Games2.898.2+95.4
Terminal-Bench Science 0.1 (Vals)Science › Multidisciplinary10.058.6+48.6
Terminal-Bench-Science 0.1Science › Multidisciplinary20.067.6+47.6
AA-PersonalityBench Informativeness indexLanguage & writing › Style & personality75.337.7−37.9
AA-3D v1.0 EloSoftware › Frontend & design9981259+35.5
AutomationBench-AA 1.0.6Tool & computer use › Customer service & operations36.071.3+35.3
Code Migration v1.1 (Vals Index v2.2 subset)Software › Software engineering50.285.3+35.1
Humanity's Last Exam DiamondKnowledge & reasoning › General knowledge29.664.0+34.4
AutomationBench 1.0.6 (Zapier)Tool & computer use › Customer service & operations12.041.3+29.2
TaxCalcBench TY25Professional › Finance, accounting & tax4.032.0+28.0
Terminal-Bench 4.0 (Artificial Analysis)Software › Software engineering32.860.1+27.3
ARC-AGI-2Knowledge & reasoning › Abstract reasoning61.385.8+24.6
Terminal-Bench 4.0Software › Software engineering36.460.6+24.2
Blueprint-Bench 2Perception & media › Spatial & visual reasoning0.30.6+24.0
LMArena Agent Arena Net Improvement (Steering Burden)Tool & computer use › General agents-0.611.3+23.2
AA-PersonalityBench Expressiveness indexLanguage & writing › Style & personality77.255.3−22.1
Artificial Analysis Intelligence Index v5.0.1Composite indexes36.557.3+20.8
AA-PersonalityBench Transparency indexLanguage & writing › Style & personality77.256.9−20.5
Artificial Analysis Strategy & Ops IndexProfessional › General knowledge work39.159.4+20.3
Ramp SWE-Bench v2 (Bash harness)Software › Software engineering52.872.4+19.6
Vals Science IndexScience › Multidisciplinary38.557.8+19.3
Artificial Analysis Engineering IndexScience › Engineering42.561.2+18.6
Artificial Analysis Legal IndexProfessional › Legal38.755.0+16.4
Artificial Analysis Healthcare & Medical IndexHealth & medicine37.553.3+15.8
Artificial Analysis Economics IndexComposite indexes42.757.4+14.7
LMArena Agent Arena Praise vs ComplaintTool & computer use › General agents-5.523.3+14.4
AA-Omniscience v1.1 IndexKnowledge & reasoning › General knowledge12.340.4+14.0
VIALS-AA (pass^3)Science › Life sciences49.162.7+13.7
Artificial Analysis Finance & Accounting IndexProfessional › Finance, accounting & tax43.657.2+13.6
CritPt v2 (agentic, avg@5)Science › Physical sciences66.378.3+12.0
AA-PersonalityBench Conviction indexLanguage & writing › Style & personality67.679.2+11.7
LiveBench 2.0 (2026-11-06 release)Composite indexes68.980.3+11.4
SimpleBenchKnowledge & reasoning › General knowledge59.269.9+10.7
AA-PersonalityBench Warmth indexLanguage & writing › Style & personality69.560.1−9.6
FrontierCode 2.0 (Artificial Analysis)Software › Software engineering34.944.4+9.4
GDP.pdf (Artificial Analysis, All-pass)Professional › General knowledge work20.830.0+9.2
Excel Modeling Benchmark v1.1Professional › Finance, accounting & tax65.574.3+8.8
ARC-AGI-1 Semi-PrivateKnowledge & reasoning › Abstract reasoning89.097.5+8.5
VIALS-AA (average of 3)Science › Life sciences60.568.9+8.5
DualEntry Accounting AI BenchmarkProfessional › Finance, accounting & tax76.284.2+8.0
HLAB-Vals-1.1Professional › Legal4.212.1+7.9
BioMysteryBench (Vals Terminus 2)Science › Life sciences70.477.8+7.4
Vals Index v2.2Composite indexes58.766.1+7.3
Legal Research BenchProfessional › Legal43.837.0−6.7
Finance Agent v2.1Professional › Finance, accounting & tax65.458.9−6.5
LMArena Agent Arena Steering BurdenTool & computer use › General agents0.813.4+6.3
LMArena Agent Arena Confirmed SuccessTool & computer use › General agents0.210.9+5.3
AA-PersonalityBench Slop indexLanguage & writing › Style & personality10.515.4+4.9
AA-PersonalityBench Organization indexLanguage & writing › Style & personality41.045.7+4.8
MultiChallengeLanguage & writing › Conversation67.772.0+4.3
AA-PersonalityBench Imagination indexLanguage & writing › Style & personality63.860.1−3.8
Vals Legal IndexProfessional › Legal70.366.7−3.6
VIALS-AA (pass@3)Science › Life sciences71.474.5+3.1
Vibe Code Bench v1.1Software › Frontend & design90.487.6−2.8
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & medicine35.037.2+2.2
AA-Briefcase v1.1 EloProfessional › General knowledge work15721612+2.2
MysteryMechanismMath35.136.9+1.8
LMArena Agent Arena Tool HallucinationTool & computer use › Tool calling-1.51.2+1.4
LMArena Agent Arena Bash ReliabilityTool & computer use › General agents6.64.3−1.1
SWE Atlas v1.1 - OverallSoftware › Software engineering59.460.5+1.0
Vals Finance IndexProfessional › Finance, accounting & tax57.456.7−0.8
Tax Agent BenchProfessional › Finance, accounting & tax62.262.9+0.7
AA-AnalystAgent (pass^5)Professional › General knowledge work53.853.80.0
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…