Atlas

Compare Models

GPT-6.2 Sol vs Grok 4.7

GPT-6.2 Sol leads by 6.2 capability points (≈95% interval +3.6 to +8.8): a clear lead over Grok 4.7 · P(higher) ≈ 100%

Wins 33 of 48 shared benchmarks (15 lost)

Capability vs. relative cost
GPT-6.2 SolGrok 4.7Closed 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 SolvsGrok 4.7capability vs. relative costData Nov 17, 2026Village LLM Atlasatlas.village.dev

Benchmark wins33 – 15 · 48 shared

Comparing matching setups only (3 benchmarks excluded) · suites counted once
← GPT-6.2 SolGrok 4.7 →

GPT-6.2 SolvsGrok 4.7wins by categoryData Nov 17, 2026Data: Artificial Analysis, Vals AI and ARC Prize FoundationVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Grok 4.7 minus GPT-6.2 Sol.

ARC-AGI-3 (Semi-Private)General QA97.47.0−90.4
CyberGym-E2E-AA (Artificial Analysis)Cybersecurity0.074.0+74.0
Terminal-Bench-Science 0.1Science66.711.4−55.2
Terminal-Bench Science 0.1 (Vals)Science58.610.0−48.6
Humanity's Last Exam DiamondGeneral QA64.023.6−40.4
Terminal-Bench 4.0 (Artificial Analysis)Code58.624.7−33.8
Code Migration v1.1 (Vals Index v2.2 subset)Code85.351.8−33.5
Terminal-Bench 4.0Code60.628.8−31.8
ARC-AGI-2General QA85.861.4−24.4
Vals Science IndexIndexes57.834.8−22.9
DeepsecBench-AA (Artificial Analysis)Cybersecurity48.926.9−22.0
Artificial Analysis Intelligence Index v5.0.1Indexes57.335.5−21.8
AA-PersonalityBench Imagination indexChat & Writing56.235.3−21.1
AA-PersonalityBench Expressiveness indexChat & Writing55.775.3+19.7
Artificial Analysis Engineering IndexIndexes61.241.8−19.4
Artificial Analysis Cyber Index v1Indexes37.756.4+18.7
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine37.221.1−16.1
AA-PersonalityBench Warmth indexChat & Writing60.544.8−15.8
FrontierCode 2.0 (Artificial Analysis)Code42.828.7−14.1
MysteryMechanismScience36.925.2−11.7
Artificial Analysis Economics IndexIndexes57.445.7−11.7
AA-PersonalityBench Transparency indexChat & Writing47.458.1+10.8
Legal Research BenchLegal37.047.1+10.1
CritPt v2 (agentic, avg@5)Science71.462.3−9.1
BioMysteryBench (Vals Terminus 2)Science77.869.3−8.5
Artificial Analysis Legal IndexIndexes55.047.2−7.8
AA-PersonalityBench Slop indexChat & Writing1.08.6+7.7
AA-PersonalityBench Informativeness indexChat & Writing25.933.4+7.6
Vals Index v2.2Indexes66.158.6−7.4
ARC-AGI-1 Semi-PrivateGeneral QA97.590.2−7.3
GDP.pdf (Artificial Analysis, All-pass)Professional Work30.023.2−6.8
Artificial Analysis Healthcare & Medical IndexIndexes53.346.6−6.7
HLAB-Vals-1.1Legal12.118.3+6.2
Tax Agent BenchFinance & Accounting62.969.0+6.1
Vals Legal IndexIndexes66.772.7+6.0
AutomationBench-AA 1.0.6Agents71.365.6−5.7
Artificial Analysis Finance & Accounting IndexIndexes57.251.7−5.5
Artificial Analysis Strategy & Ops IndexIndexes59.454.8−4.6
AA-Omniscience v1.1 IndexIndexes40.431.2−4.6
CWE-Bench-AA (Artificial Analysis)Cybersecurity64.268.3+4.2
AA-Briefcase v1.1 EloAgents15771636+3.2
AA-PersonalityBench Conviction indexChat & Writing85.782.9−2.8
Excel Modeling Benchmark v1.1Finance & Accounting74.371.5−2.8
AA-AnalystAgent (pass^5)Professional Work52.555.0+2.5
Vals Finance IndexIndexes56.758.8+2.2
Finance Agent v2.1Finance & Accounting58.960.6+1.7
Vibe Code Bench v1.1Code87.686.2−1.4
AA-PersonalityBench Organization indexChat & Writing30.829.6−1.3

GPT-6.2 SolvsGrok 4.7normalised score marginsData Nov 17, 2026Data: Artificial Analysis, Vals AI and ARC Prize FoundationVillage 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…