Atlas

Compare Models

Gemini 3.7 Flash vs Inkling 2 Small

Inkling 2 Small leads by 0.8 capability points (≈95% interval −1.2 to +2.7): statistically tied · P(higher) ≈ 77%

Wins 30 of 47 shared benchmarks (16 lost, 1 tied)

Capability vs. relative cost
Gemini 3.7 FlashInkling 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.

Gemini 3.7 FlashvsInkling 2 Smallcapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins16 – 30 · 47 shared, 1 tied

Comparing matching setups only (51 benchmarks excluded) · suites counted once
← Gemini 3.7 FlashInkling 2 Small →

Gemini 3.7 FlashvsInkling 2 Smallwins by categoryData Nov 15, 2026Data: Vals AI, Artificial Analysis, Mercor and 9 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 Gemini 3.7 Flash.

Terminal-Bench 4.0 — SecurityCode33.30.0−33.3
CyberBench v1.1 OverallCybersecurity43.872.7+29.0
AA-Briefcase v1.1 EloAgents11151527+22.6
Legal Research Bench Civil LitigationLegal29.651.9+22.2
Legal Research Bench HealthLegal50.070.0+20.0
APEX-Agents 1.1Professional Work79.460.9−18.5
Legal Research Bench Administrative/RegulatoryLegal42.460.6+18.2
EnterpriseBench: CoreCraftProfessional Work55.971.8+15.9
Terminal-Bench 4.0 — HardwareCode13.30.0−13.3
Vibe Code Bench v1.1Code70.481.5+11.2
Terminal-Bench 4.0 — SoftwareCode7.418.5+11.1
Terminal-Bench 4.0 — OperationsCode14.83.7−11.1
Arena Code WebDev — ReactCode15841659+10.1
GDPval-AA v2.1Professional Work14041630+10.0
Legal Research Bench Business/CommercialLegal37.246.5+9.3
Arena Code WebDev — HTMLCode15931660+9.3
Legal Research Bench FamilyLegal13.622.7+9.1
Legal Research Bench ImmigrationLegal36.427.3−9.1
Terminal-Bench 4.0 — MLCode12.121.2+9.1
DAYJOB: Finance (October 2026 rescoring)Finance & Accounting5.314.3+9.0
AA-AnalystAgent (pass^5)Professional Work60.051.3−8.8
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine12.220.6+8.3
Tax Agent BenchFinance & Accounting57.765.8+8.1
Legal Research Bench Constitutional/Civil RightsLegal31.639.5+7.9
SciCode 1.0.1 (Artificial Analysis)Code59.852.3−7.5
DeepSWE v1.1Code65.572.1+6.6
Humanity's Sixth SenseMultimodal39.833.3−6.5
APEX-Accounting 1.1Finance & Accounting57.262.0+4.8
Terminal-Bench 4.0 — ScienceCode11.916.7+4.8
Terminal-Bench 2.1Code85.881.3−4.5
Terminal-Bench Science 0.1 (Vals)Science7.12.9−4.3
Legal Research Bench Weighted PassLegal84.187.9+3.7
ChartographyMultimodal40.436.7−3.7
EnterpriseOps-Gym-AAAgents50.446.8−3.6
Legal Research Bench CriminalLegal33.336.8+3.5
ARC-AGI-2General QA84.681.7−2.9
ATLAS-Finance v1.1Finance & Accounting3.36.0+2.7
Humanity's Last Exam DiamondGeneral QA31.228.7−2.5
ComplexConstraintsChat & Writing42.444.5+2.1
GDP.pdf (Artificial Analysis, All-pass)Professional Work23.625.2+1.6
SWE-bench MultilingualCode82.681.2−1.4
SWE-bench VerifiedCode81.082.2+1.2
ARC-AGI-1 Semi-PrivateGeneral QA95.596.5+1.0
Terminal-Bench 4.0 (Artificial Analysis)Code13.613.1−0.5
ProgramBenchCode0.00.5+0.5
CharXiv Reasoning (with tools)Multimodal88.789.1+0.4
Terminal-Bench 4.0 — MediaCode0.00.00.0

Gemini 3.7 FlashvsInkling 2 Smallnormalised score marginsData Nov 15, 2026Data: Vals AI, Artificial Analysis, Mercor and 9 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…