Atlas

Compare Models

GPT-6.1 Terra vs GPT-6.2 Sol

GPT-6.2 Sol leads by 1.7 capability points (≈95% interval −0.9 to +4.3): statistically tied · P(higher) ≈ 90%

Wins 54 of 63 shared benchmarks (8 lost, 1 tied)

Capability vs. relative cost
GPT-6.1 TerraGPT-6.2 SolClosed weightsOpen weights

Benchmark wins8 – 54 · 63 shared, 1 tied

Comparing matching setups only (2 benchmarks excluded) · suites counted once
← GPT-6.1 TerraGPT-6.2 Sol →
Data: Artificial Analysis, Vals AI, Arena 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 GPT-6.1 Terra.

KernelGen1P - OpenAI release percentageSoftware › Data & ML45.374.7+53.9
FrontierCUA v1 pass@1Tool & computer use › Computer & browser use7.525.0+17.5
Humanity's Last Exam DiamondKnowledge & reasoning › General knowledge47.364.0+16.7
PostTrainBench Lite - release-reported performanceSoftware › Data & ML60.544.9−15.6
Blueprint-Bench 2Perception & media › Spatial & visual reasoning0.40.6+14.9
Bug Hunt BenchSoftware › Software engineering38.752.1+13.3
Code Migration v1.1 (Vals Index v2.2 subset)Software › Software engineering72.285.3+13.2
ChartographyPerception & media › Documents & charts63.275.1+11.9
AA-Briefcase v1.1 EloProfessional › General knowledge work14001612+11.6
Terminal-Bench 4.0Software › Software engineering49.560.6+11.1
AutomationBench-AA 1.0.6Tool & computer use › Customer service & operations60.271.3+11.1
Legal Research BenchProfessional › Legal26.037.0+11.1
RWE-benchScience › Engineering19.030.0+11.0
HLAB-Vals-1.1Professional › Legal1.712.1+10.4
Terminal-Bench-Science 0.1Science › Multidisciplinary57.667.6+10.0
LMArena Code Arena WebDevSoftware › Frontend & design16821757+9.9
TxBench Preclinical Pharmacology V1 Clean RerunsScience › Life sciences55.765.3+9.7
Artificial Analysis Strategy & Ops IndexProfessional › General knowledge work50.059.4+9.4
Artificial Analysis Economics IndexComposite indexes48.557.4+8.9
Artificial Analysis Legal IndexProfessional › Legal46.455.0+8.6
Artificial Analysis Healthcare & Medical IndexHealth & medicine44.753.3+8.6
Artificial Analysis Finance & Accounting IndexProfessional › Finance, accounting & tax49.057.2+8.2
TaxCalcBench TY25Professional › Finance, accounting & tax24.032.0+8.0
Vals Legal IndexProfessional › Legal58.866.7+7.9
Tax Agent BenchProfessional › Finance, accounting & tax55.062.9+7.9
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & medicine30.637.2+6.7
Artificial Analysis Engineering IndexScience › Engineering54.761.2+6.4
CWE-bench v2 Pass@1Software › Cybersecurity60.066.0+6.0
DualEntry Accounting AI BenchmarkProfessional › Finance, accounting & tax78.284.2+6.0
Artificial Analysis Intelligence Index v5.0.1Composite indexes51.357.3+5.9
Vals Index v2.2Composite indexes60.566.1+5.5
AA-Omniscience v1.1 IndexKnowledge & reasoning › General knowledge29.640.4+5.4
Gray Swan IPI Attack Success (k=15)Safety & alignment7.32.3+5.0
MentalHealthBenchHealth & medicine55.550.9−4.6
Terminal-Bench 4.0 (Artificial Analysis)Software › Software engineering55.660.1+4.5
LMArena Agent Arena Bash ReliabilityTool & computer use › General agents12.74.3−4.2
DrugDiscoveryBenchScience › Life sciences58.162.2+4.1
Vibe Code Bench v1.1Software › Frontend & design84.087.6+3.6
Vals Finance IndexProfessional › Finance, accounting & tax53.356.7+3.3
VIALS-AA (pass@3)Science › Life sciences71.474.5+3.1
Ramp SWE-Bench v2Software › Software engineering70.473.4+3.0
Finance Agent v2.1Professional › Finance, accounting & tax56.058.9+3.0
ExploitBench - release-reported exploit progressSoftware › Cybersecurity97.1100.0+2.9
VIALS-AA (average of 3)Science › Life sciences66.068.9+2.9
CritPt v2 (agentic, avg@5)Science › Physical sciences81.478.6−2.9
SWE Atlas v1.1 - OverallSoftware › Software engineering57.760.5+2.8
LMArena Agent Arena Praise vs ComplaintTool & computer use › General agents17.923.3+2.7
VIALS-AA (pass^3)Science › Life sciences60.262.7+2.5
GPQA DiamondScience › Multidisciplinary93.696.0+2.4
SimpleBenchKnowledge & reasoning › General knowledge72.269.9−2.3
MultiChallengeLanguage & writing › Conversation74.272.0−2.2
FrontierCode 2.0 (Artificial Analysis)Software › Software engineering42.344.4+2.1
LiveBench 2.0 (2026-11-06 release)Composite indexes78.680.3+1.7
OpenAI research debugging - internal release scoreSoftware › Data & ML76.474.7−1.7
LMArena Agent Arena Net Improvement (Steering Burden)Tool & computer use › General agents10.511.3+1.6
GDP.pdf (Artificial Analysis, All-pass)Professional › General knowledge work28.630.0+1.4
HealthBench ProfessionalHealth & medicine64.265.5+1.3
LMArena Agent Arena Confirmed SuccessTool & computer use › General agents8.510.9+1.2
Excel Modeling Benchmark v1.1Professional › Finance, accounting & tax73.374.3+1.0
Agents' Last Exam v1.1 (ALE v1.1)Professional › General knowledge work41.041.8+0.8
LMArena Agent Arena Steering BurdenTool & computer use › General agents13.213.4+0.1
Artificial Analysis Coding Agent Index v2.1Software › Software engineering48.648.60.0
LMArena Agent Arena Tool HallucinationTool & computer use › Tool calling1.21.20.0
Data: Artificial Analysis, Vals AI, Arena 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…