Atlas

Compare Models

GPT-6.2 Sol vs GPT-6 Astra

GPT-6 Astra leads by 0.3 capability points (≈95% interval −2.5 to +3.1): statistically tied · P(higher) ≈ 59%

Wins 29 of 52 shared benchmarks (22 lost, 1 tied)

Capability vs. relative cost
GPT-6.2 SolGPT-6 AstraClosed weightsOpen weights

Each line runs through one model’s priced reasoning settings, cheapest to most expensive; grey points are other models’ priced settings in the same range. Results whose reasoning effort the source didn’t report are left out wherever a model has named effort levels. Capability and relative cost are the published estimates (relative cost is cost per task as a multiple of GPT-5 (high)); hover a setting, or use the arrow keys, to set it beside the other model’s same setting.

GPT-6.2 SolvsGPT-6 Astracapability vs. relative costData Nov 17, 2026Village LLM Atlasatlas.village.dev

Benchmark wins22 – 29 · 52 shared, 1 tied

Comparing matching setups only (4 benchmarks excluded) · suites counted once
← GPT-6.2 SolGPT-6 Astra →

GPT-6.2 SolvsGPT-6 Astrawins by categoryData Nov 17, 2026Data: Artificial Analysis, Vals AI, OpenAI and 2 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is GPT-6 Astra minus GPT-6.2 Sol.

AA-PersonalityBench Transparency indexChat & Writing47.494.3+47.4
AA-PersonalityBench Conviction indexChat & Writing85.744.8−41.3
AA-PersonalityBench Organization indexChat & Writing30.869.5+39.1
PostTrainBench Lite - release-reported performanceCode44.977.0+32.1
ExploitBench - Internal Port (June-August 2026)Cybersecurity50.529.0−21.5
AA-PersonalityBench Imagination indexChat & Writing56.277.2+21.2
MysteryMechanismScience36.953.2+16.2
AA-PersonalityBench Slop indexChat & Writing1.016.2+15.4
AA-PersonalityBench Warmth indexChat & Writing60.575.3+14.9
KernelGen1P - OpenAI release percentageCode74.766.7−14.7
DeepsecBench-AA (Artificial Analysis)Cybersecurity51.636.9−14.6
ARC-AGI-2General QA85.895.0+9.2
Terminal-Bench-Science 0.1Science67.659.5−8.1
MentalHealthBenchSafety50.958.7+7.8
Vals Science IndexIndexes57.865.1+7.3
Code Migration v1.1 (Vals Index v2.2 subset)Code85.378.0−7.3
SecureSWE (OpenAI internal)Cybersecurity55.648.7−6.9
CritPt v2 (agentic, avg@5)Science78.384.0+5.7
HLAB-Vals-1.1Legal12.17.1−5.0
PatchBench (OpenAI internal)Cybersecurity75.670.9−4.7
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine37.241.7+4.4
Artificial Analysis Cyber Index v1Indexes37.733.4−4.3
Terminal-Bench Science 0.1 (Vals)Science58.662.9+4.3
LifeVal (GPT-6.2 launch chart)Agents83.879.8−4.0
AA-PersonalityBench Informativeness indexChat & Writing25.929.6+3.7
Terminal-Bench 4.0 (Artificial Analysis)Code56.159.6+3.5
OpenAI research debugging - internal release scoreCode74.778.0+3.4
SEC-Bench Pro - release-reported cohortCybersecurity88.385.4−2.9
AutomationBench-AA 1.0.6Agents71.368.5−2.8
HealthBench ProfessionalHealth & Medicine65.568.2+2.7
GDPval v2 (OpenAI internal)Professional Work46.944.2−2.7
AA-AnalystAgent (pass^5)Professional Work53.851.3−2.5
AA-PersonalityBench Expressiveness indexChat & Writing55.758.1+2.4
Legal Research BenchLegal37.039.4+2.4
AA-Briefcase v1.1 EloAgents16121569−2.4
GDP.pdf (Artificial Analysis, All-pass)Professional Work30.032.2+2.2
AA-Omniscience v1.1 IndexIndexes40.444.6+2.1
FrontierCode 2.0 (Artificial Analysis)Code44.442.4−2.0
Vals Legal IndexIndexes66.768.4+1.7
CWE-Bench-AA (Artificial Analysis)Cybersecurity61.763.3+1.7
Excel Modeling Benchmark v1.1Finance & Accounting74.372.7−1.6
Finance Agent v2.1Finance & Accounting58.960.4+1.5
BioMysteryBench (Vals Terminus 2)Science77.879.3+1.5
Vals Finance IndexIndexes56.755.6−1.0
Terminal-Bench 4.0Code60.659.6−1.0
ARC-AGI-1 Semi-PrivateGeneral QA97.598.5+1.0
Tax Agent BenchFinance & Accounting62.963.3+0.4
Artificial Analysis Intelligence Index v5.0.1Indexes57.356.9−0.4
ARC-AGI-3 (Semi-Private)General QA98.298.6+0.3
Humanity's Last Exam DiamondGeneral QA64.063.8−0.2
AutomationBench 1.0.6 (Zapier)Agents41.341.4+0.1
CyberGym-E2E-AA (Artificial Analysis)Cybersecurity0.00.00.0

GPT-6.2 SolvsGPT-6 Astranormalised score marginsData Nov 17, 2026Data: Artificial Analysis, Vals AI, OpenAI and 2 othersVillage LLM Atlasatlas.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…