Atlas

Compare Models

Gemini 4 Argon vs GPT-6.2 Sol

GPT-6.2 Sol leads by 1.3 capability points (≈95% interval −1.3 to +3.9): statistically tied · P(higher) ≈ 84%

Wins 40 of 75 shared benchmarks (35 lost)

Capability vs. relative cost
Gemini 4 ArgonGPT-6.2 SolClosed weightsOpen weights

Benchmark wins35 – 40 · 75 shared

Comparing matching setups only (36 benchmarks excluded) · suites counted once
← Gemini 4 ArgonGPT-6.2 Sol →
Data: Artificial Analysis, Vals AI, Surge AI and 15 othersatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is GPT-6.2 Sol minus Gemini 4 Argon.

AA-PersonalityBench Slop indexLanguage & writing › Style & personality67.615.4−52.8
AA-PersonalityBench Informativeness indexLanguage & writing › Style & personality82.937.7−45.6
KradleBenchGames28.667.9+39.3
AA-PersonalityBench Transparency indexLanguage & writing › Style & personality21.956.9+35.3
AA-PersonalityBench Expressiveness indexLanguage & writing › Style & personality84.855.3−29.8
DAYJOB: Law (November 2026 release)Professional › Legal5.932.7+26.8
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & medicine13.937.2+23.3
SimpleBenchKnowledge & reasoning › General knowledge91.769.9−21.8
sudo L7 (October 2026 rescoring)Software › Software engineering16.136.1+20.0
AA-PersonalityBench Warmth indexLanguage & writing › Style & personality79.160.1−19.2
Legal Research BenchProfessional › Legal54.837.0−17.8
Ramp SWE-Bench v2 (Bash harness)Software › Software engineering54.872.4+17.6
FrontierCode 2.0 (Artificial Analysis)Software › Software engineering25.041.4+16.4
Tax Agent BenchProfessional › Finance, accounting & tax79.062.9−16.1
GDP.xlsxProfessional › General knowledge work38.322.3−16.0
AA-PersonalityBench Conviction indexLanguage & writing › Style & personality63.879.2+15.6
AA-3D v1.0 EloSoftware › Frontend & design11471259+15.3
HLAB-Vals-1.1Professional › Legal27.112.1−15.0
Terminal-Bench Science 0.1 (Vals)Science › Multidisciplinary44.358.6+14.3
AA-PersonalityBench Imagination indexLanguage & writing › Style & personality46.760.1+13.5
Terminal-Bench-Science 0.1Science › Multidisciplinary48.660.5+11.9
Code Migration v1.1 (Vals Index v2.2 subset)Software › Software engineering74.485.3+10.9
AA-PersonalityBench Organization indexLanguage & writing › Style & personality56.245.7−10.6
Vals Legal IndexProfessional › Legal76.866.7−10.1
DAYJOB: AccountingProfessional › Finance, accounting & tax18.027.6+9.6
SWE Atlas v1.1 - OverallSoftware › Software engineering70.160.5−9.6
Vals Finance IndexProfessional › Finance, accounting & tax66.056.7−9.3
AutomationBench 1.0.6 (Zapier)Tool & computer use › Customer service & operations50.141.3−8.8
Finance Agent v2.1Professional › Finance, accounting & tax67.758.9−8.7
MysteryMechanismMath45.536.9−8.6
Dual-Arm Sim (task completion)Tool & computer use › Robotics11.33.8−7.5
ARC-AGI-3 (Semi-Private)Games97.390.1−7.2
DualEntry Accounting AI BenchmarkProfessional › Finance, accounting & tax91.184.2−6.9
Artificial Analysis Intelligence Index v5.0.1Composite indexes50.457.3+6.8
MultiChallengeLanguage & writing › Conversation78.772.0−6.7
DAYJOB: Healthcare (October 2026 rescoring)Health & medicine23.617.2−6.4
AutomationBench-AA 1.0.6Tool & computer use › Customer service & operations77.571.3−6.2
Professional Reasoning Benchmark - LegalProfessional › Legal49.443.3−6.1
RiemannBench (November 2026 rescoring)Math74.480.0+5.6
ARC-AGI-2Knowledge & reasoning › Abstract reasoning91.385.8−5.4
LMArena Agent Arena Bash ReliabilityTool & computer use › General agents-6.04.3+5.1
CritPt v2 (agentic, avg@5)Science › Physical sciences72.977.7+4.9
GDP.pdf (November 2026 rescoring)Professional › General knowledge work53.648.8−4.8
GDP.pdf (Artificial Analysis, All-pass)Professional › General knowledge work25.630.0+4.4
GraphWalks BFS: 256K to 1M AggregateKnowledge & reasoning › Reading & long context84.280.1−4.1
LVBenchPerception & media › Video understanding91.787.6−4.1
Agents' Last Exam v1.1 ScoreProfessional › General knowledge work69.073.0+4.0
AA-AnalystAgent (pass^5)Professional › General knowledge work48.852.5+3.8
ChartographyPerception & media › Documents & charts71.675.1+3.5
Humanity's Last Exam DiamondKnowledge & reasoning › General knowledge53.256.5+3.3
DrugDiscoveryBenchScience › Life sciences65.462.2−3.2
VIALS-AA (pass^3)Science › Life sciences59.662.7+3.1
AA-Briefcase v1.1 EloProfessional › General knowledge work14851541+3.1
Terminal-Bench 4.0Software › Software engineering57.660.6+3.0
AA-Omniscience v1.1 IndexKnowledge & reasoning › General knowledge46.140.4−2.9
VIALS-AA (average of 3)Science › Life sciences66.368.9+2.7
LiveBench 2.0 (2026-11-06 release)Composite indexes77.780.3+2.6
Blueprint-Bench 2Perception & media › Spatial & visual reasoning0.50.6+2.6
Vals Science IndexScience › Multidisciplinary55.457.8+2.4
DAYJOB: Finance (October 2026 rescoring)Professional › Finance, accounting & tax30.532.8+2.3
HANDBOOK.md Agents v1.1Tool & computer use › Customer service & operations36.634.5−2.1
LMArena Agent Arena Net Improvement (Steering Burden)Tool & computer use › General agents10.311.3+2.0
VIALS-AA (pass@3)Science › Life sciences72.774.5+1.9
DAYJOB: TaxProfessional › Finance, accounting & tax3.24.8+1.6
BioMysteryBench (Vals Terminus 2)Science › Life sciences76.377.8+1.5
Humanity's Sixth SensePerception & media › Spatial & visual reasoning55.456.8+1.4
Terminal-Bench 4.0 (Artificial Analysis)Software › Software engineering55.156.1+1.0
ARC-AGI-1 Semi-PrivateKnowledge & reasoning › Abstract reasoning98.597.5−1.0
LMArena Agent Arena Confirmed SuccessTool & computer use › General agents12.610.9−0.9
ComplexConstraints (November 2026 rescoring)Language & writing › Instruction following70.369.5−0.8
TxBench Preclinical Pharmacology V1 Clean RerunsScience › Life sciences64.765.3+0.7
LMArena Agent Arena Steering BurdenTool & computer use › General agents12.213.4+0.6
Excel Modeling Benchmark v1.1Professional › Finance, accounting & tax74.874.3−0.5
LMArena Agent Arena Praise vs ComplaintTool & computer use › General agents23.823.3−0.2
LMArena Agent Arena Tool HallucinationTool & computer use › Tool calling0.81.2+0.2
Data: Artificial Analysis, Vals AI, Surge AI and 15 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…