Atlas

Compare Models

GPT-6 Sol vs Inkling 2 Small

GPT-6 Sol leads by 3.6 capability points (≈95% interval +1.5 to +5.8): a clear lead over Inkling 2 Small · P(higher) ≈ 100%

Wins 28 of 46 shared benchmarks (17 lost, 1 tied)

Capability vs. relative cost
GPT-6 SolInkling 2 SmallClosed 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 SolvsInkling 2 Smallcapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins28 – 17 · 46 shared, 1 tied

Comparing matching setups only (54 benchmarks excluded) · suites counted once
← GPT-6 SolInkling 2 Small →

GPT-6 SolvsInkling 2 Smallwins by categoryData Nov 15, 2026Data: Vals AI, Artificial Analysis, Mercor and 8 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Inkling 2 Small minus GPT-6 Sol.

ArXivMath 08/2026 (Native Harnesses)Math93.036.8−56.1
Terminal-Bench 4.0Code49.412.6−36.8
Legal Research Bench Civil LitigationLegal18.551.9+33.3
Terminal-Bench 4.0 (Artificial Analysis)Code43.913.1−30.8
EnterpriseBench: CoreCraftProfessional Work43.171.8+28.7
SRE Bench — Fully SolvedCode30.51.9−28.6
Terminal-Bench Science 0.1 (Vals)Science30.02.9−27.1
CritPt v2 (agentic, avg@5)Science80.955.7−25.1
Terminal-Bench-Science 0.1Science30.05.7−24.3
Legal Research Bench Administrative/RegulatoryLegal39.460.6+21.2
Legal Research Bench Constitutional/Civil RightsLegal18.439.5+21.1
Legal Research Bench Business/CommercialLegal27.946.5+18.6
ChartographyMultimodal53.636.7−16.9
Humanity's Last Exam DiamondGeneral QA45.228.7−16.5
BrokenArXiv 08/2026 (Native Harnesses)Math78.663.1−15.5
AA-3D v1.0 EloCode11311027−15.1
Legal Research Bench FamilyLegal9.122.7+13.6
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine33.920.6−13.3
Tax Agent BenchFinance & Accounting53.065.8+12.7
Vals Science IndexIndexes45.032.7−12.3
Legal Research Bench CriminalLegal26.336.8+10.5
FrontierCode 2.0 (Artificial Analysis)Code35.625.4−10.2
Legal Research Bench HealthLegal60.070.0+10.0
Legal Research Bench ImmigrationLegal36.427.3−9.1
Legal Research Bench Weighted PassLegal79.287.9+8.7
ARC-AGI-2General QA89.681.7−7.9
APEX-Agents 1.1Professional Work68.760.9−7.8
ComplexConstraintsChat & Writing50.844.5−6.3
Vibe Code Bench v1.1Code87.881.5−6.3
DeepSWE v1.1Code66.672.1+5.5
SciCode 1.0.1 (Artificial Analysis)Code57.652.3−5.3
CyberBench v1.1 OverallCybersecurity78.072.7−5.2
DAYJOB: Finance (October 2026 rescoring)Finance & Accounting19.514.3−5.2
GDPval-AA v2.1Professional Work15141630+5.1
MysteryMechanismScience30.225.2−5.0
BioMysteryBench (Vals Terminus 2)Science74.870.0−4.8
Terminal-Bench 2.1Code85.481.3−4.1
AA-Briefcase v1.1 EloAgents14761527+2.8
AA-Omniscience v1.1 IndexIndexes21.416.1−2.6
APEX-Accounting 1.1Finance & Accounting59.762.0+2.3
Humanity's Sixth SenseMultimodal31.233.3+2.1
SWE-bench VerifiedCode80.382.2+1.9
ProgramBenchCode2.00.5−1.5
ARC-AGI-1 Semi-PrivateGeneral QA95.596.5+1.0
SWE-bench MultilingualCode81.781.2−0.5
GDP.pdf (Artificial Analysis, All-pass)Professional Work25.225.20.0

GPT-6 SolvsInkling 2 Smallnormalised score marginsData Nov 15, 2026Data: Vals AI, Artificial Analysis, Mercor and 8 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…