Atlas

Compare Models

Claude Sonnet 5.5 vs GPT-6.2 Sol

GPT-6.2 Sol leads by 0.4 capability points (≈95% interval −2.3 to +3.1): statistically tied · P(higher) ≈ 61%

Wins 37 of 69 shared benchmarks (32 lost)

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

Benchmark wins32 – 37 · 69 shared

Comparing matching setups only (33 benchmarks excluded) · suites counted once
← Claude Sonnet 5.5GPT-6.2 Sol →
Data: Artificial Analysis, Surge AI, Arena and 16 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 Sonnet 5.5.

AA-PersonalityBench Informativeness indexLanguage & writing › Style & personality84.837.7−47.5
AA-PersonalityBench Transparency indexLanguage & writing › Style & personality10056.9−43.5
AA-PersonalityBench Organization indexLanguage & writing › Style & personality84.845.7−39.5
EnterpriseBench: CoreCraftTool & computer use › Customer service & operations70.836.4−34.4
AA-PersonalityBench Imagination indexLanguage & writing › Style & personality84.860.1−24.9
ARC-AGI-3 (Semi-Private)Games72.497.4+24.9
AA-PersonalityBench Conviction indexLanguage & writing › Style & personality58.179.2+21.3
AA-PersonalityBench Expressiveness indexLanguage & writing › Style & personality73.355.3−18.2
Blueprint-Bench 2Perception & media › Spatial & visual reasoning0.40.6+17.7
DAYJOB: Healthcare (October 2026 rescoring)Health & medicine34.017.2−16.8
Ramp SWE-Bench v2 (Bash harness)Software › Software engineering57.672.4+14.8
sudo L7 (October 2026 rescoring)Software › Software engineering50.636.1−14.5
Terminal-Bench-Science 0.1Science › Multidisciplinary53.367.6+14.3
TaxCalcBench TY25Professional › Finance, accounting & tax18.032.0+14.0
ChartographyPerception & media › Documents & charts61.675.1+13.5
AA-PersonalityBench Warmth indexLanguage & writing › Style & personality73.360.1−13.4
HANDBOOK.md Agents v1.1Tool & computer use › Customer service & operations47.734.5−13.2
Agents' Last Exam v1.1 ScoreProfessional › General knowledge work60.473.0+12.6
AA-PersonalityBench Slop indexLanguage & writing › Style & personality27.715.4−12.4
MysteryMechanismMath49.136.9−12.2
Terminal-Bench Science 0.1 (Vals)Science › Multidisciplinary47.158.6+11.4
AA-Briefcase v1.1 EloProfessional › General knowledge work18151612−11.1
Humanity's Last Exam DiamondKnowledge & reasoning › General knowledge53.364.0+10.7
DeepsecBench — Revalidated V1 (233 issues)Software › Cybersecurity37.547.3+9.8
LMArena Agent Arena Net Improvement (Steering Burden)Tool & computer use › General agents15.611.3−8.3
SWE Atlas v1.1 - OverallSoftware › Software engineering68.360.5−7.8
DAYJOB: Finance (October 2026 rescoring)Professional › Finance, accounting & tax40.532.8−7.7
DAYJOB: AccountingProfessional › Finance, accounting & tax35.227.6−7.6
AA-3D v1.0 EloSoftware › Frontend & design12041259+7.6
GDP.xlsxProfessional › General knowledge work29.122.3−6.8
DAYJOB: TaxProfessional › Finance, accounting & tax10.84.8−6.0
SimpleBenchKnowledge & reasoning › General knowledge75.969.9−6.0
DAYJOB: Law (November 2026 release)Professional › Legal26.832.7+5.9
AA-Omniscience v1.1 IndexKnowledge & reasoning › General knowledge30.540.4+5.0
Vibe Code Bench v1.1Software › Frontend & design92.487.6−4.8
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & medicine41.737.2−4.4
GDP.pdf (Artificial Analysis, All-pass)Professional › General knowledge work25.830.0+4.2
DualEntry Accounting AI BenchmarkProfessional › Finance, accounting & tax80.284.2+4.0
LMArena Agent Arena Bash ReliabilityTool & computer use › General agents11.94.3−3.8
Dual-Arm Sim (task completion)Tool & computer use › Robotics0.03.8+3.8
AA-AnalystAgent (pass^5)Professional › General knowledge work57.553.8−3.7
VIALS-AA (pass@3)Science › Life sciences70.874.5+3.7
LMArena Agent Arena Praise vs ComplaintTool & computer use › General agents30.723.3−3.7
Terminal-Bench 4.0 (Artificial Analysis)Software › Software engineering63.660.1−3.5
LiveBench 2.0 (2026-11-06 release)Composite indexes76.880.3+3.5
VIALS-AA (average of 3)Science › Life sciences65.468.9+3.5
LMArena Agent Arena Confirmed SuccessTool & computer use › General agents17.610.9−3.3
Artificial Analysis Economics IndexComposite indexes54.157.4+3.3
RiemannBench (November 2026 rescoring)Math76.880.0+3.2
VIALS-AA (pass^3)Science › Life sciences59.662.7+3.1
CritPt v2 (agentic, avg@5)Science › Physical sciences75.778.6+2.9
Artificial Analysis Legal IndexProfessional › Legal52.255.0+2.9
Artificial Analysis Intelligence Index v5.0.1Composite indexes54.557.3+2.7
FrontierCode 2.0 (Artificial Analysis)Software › Software engineering41.744.4+2.7
Artificial Analysis Engineering IndexScience › Engineering58.561.2+2.7
GDP.pdf (November 2026 rescoring)Professional › General knowledge work51.248.8−2.4
ComplexConstraints (November 2026 rescoring)Language & writing › Instruction following67.269.5+2.3
ARC-AGI-2Knowledge & reasoning › Abstract reasoning83.985.8+1.9
HLAB-Vals-1.1Professional › Legal10.412.1+1.7
Excel Modeling Benchmark v1.1Professional › Finance, accounting & tax75.974.3−1.6
Artificial Analysis Finance & Accounting IndexProfessional › Finance, accounting & tax55.857.2+1.4
Artificial Analysis Strategy & Ops IndexProfessional › General knowledge work58.159.4+1.3
Gray Swan IPI Attack Success (k=15)Safety & alignment3.42.3+1.1
ARC-AGI-1 Semi-PrivateKnowledge & reasoning › Abstract reasoning96.597.5+1.0
TxBench Preclinical Pharmacology V1 Clean RerunsScience › Life sciences64.365.3+1.0
Artificial Analysis Healthcare & Medical IndexHealth & medicine54.253.3−0.9
LMArena Agent Arena Steering BurdenTool & computer use › General agents14.813.4−0.7
AutomationBench-AA 1.0.6Tool & computer use › Customer service & operations71.871.3−0.5
LMArena Agent Arena Tool HallucinationTool & computer use › Tool calling1.01.2+0.1
Data: Artificial Analysis, Surge AI, Arena and 16 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…