Skip to main content
CARLA HQ
MODEL PROFILE & BENCHMARKSRank #2 Global EloOpen Source (Apache 2.0)

Kumo Tabular Small (NVIDIA / Kumo AI)

Kumo AI and NVIDIA's in-context tabular foundation model, pre-trained on diverse structured datasets for zero-shot classification and quantile regression.

Tournament Elo Rating
1042±50 (95% CI)
Global Rank #2 of 5
Overall Improvability
1.23%
Class: 1.13% • Regr: 1.53%
Tournament Matchup Record
29W - 21L(34T)
Across all pairwise combinations
Mean Latency & Runtime
14.52s
Native WebGPU & WASM Browser Runtime (via ONNX Runtime)
EMPIRICAL BENCHMARK MATRIX

Kumo Tabular Small Dataset-by-Dataset Telemetry

Results on deterministic IID or non-IID evaluation splits, secondary metrics, suite ceilings, and per-dataset Improvability regret across all 21 benchmark datasets.

DatasetDomainTaskRowsCols
Splits?Repeats × folds from the dataset's persisted evaluation protocol. The same row indices are used for every model.
Score?Primary benchmark metric: 1 - AUC error loss for binary classification, Log-Loss for multiclass, RMSE for regression (lower is better for all three). Evaluated out-of-sample on persisted, model-identical split indices.
Ceiling?Best foundation model benchmark score: lowest 1 - AUC error loss for binary classification, lowest Log-Loss for multiclass, lowest RMSE for regression achieved across the suite.
Improv.?TabArena normalized error regret: (Loss - Ceiling) / (Dummy - Ceiling). 0.0% = Holds suite ceiling.
Latency?Mean inference time per persisted split running out-of-sample inference.
Sheet
AbaloneAgriculture, Forestry & FishingREG4,17783×32.022.020.0%1.8s
AdultEconomics & Public PolicyBIN48,842143×30.06770.06770.0%96.9s
Airfoil Self NoiseUCIREG1,503510×31.021.020.0%726ms
Amazon Employee AccessInformation Technology & Enterprise SecurityBIN32,76993×30.13920.1232+4.2%48.4s
Bank MarketingFinance & BankingBIN45,211153×30.18800.1840+1.3%85.0s
Blood Transfusion Service CenterHealthcare & BiomedicineBIN748420×30.24520.2445+0.3%501ms
Breast WHealthcare & Life SciencesBIN699920×30.00550.0052+0.1%534ms
CarAutomotive & Fleet ManagementMULTI1,728610×30.02530.0108+1.1%653ms
Compas Two YearsLegal & Public SafetyBIN5,278133×30.26170.2612+0.2%2.7s
Credit GFinance & BankingBIN1,0002010×30.19560.1938+0.6%775ms
DiabetesHealthcare & BiomedicineBIN768810×30.16210.1607+0.4%545ms
Employee SalariesHuman Resources & Workforce AnalyticsREG9,228103×39,0297,427+7.4%5.5s
Fitness ClubFitness, Sports & RecreationBIN1,500610×30.17920.17920.0%756ms
House SalesReal Estate & Property ValuationREG21,613203×398,51298,5120.0%26.7s
HousesReal Estate & Urban PlanningREG20,64083×337,67037,481+0.2%19.7s
Monks Problems 2Data Science & Artificial IntelligenceBIN601620×30.00000.00000.0%503ms
PhonemeSpeech Processing & Acoustic EngineeringBIN5,40453×30.02460.0202+0.9%2.3s
SpambaseCybersecurity & IT InfrastructureBIN4,601573×30.00730.0066+0.2%5.0s
Telco Customer ChurnTelecommunications & Subscription ServicesBIN7,043203×30.15950.1474+3.4%4.7s
TitanicMaritime Safety & Actuarial Risk AnalysisBIN1,3091110×30.12760.1146+3.4%837ms
VehicleAutomotive Engineering & Computer VisionMULTI8461810×30.22440.1986+2.2%528ms
BROWSER DEPLOYMENT FEASIBILITY

Which Tabular Foundation Model Can Actually Run in the Browser?

Evaluating TabICLv2, TabPFN-3, Google TabFM, and EXAONE-Tabular on parameter size, VRAM footprint, licensing terms, and WebGPU client-side execution compatibility.

STATISTICAL METHODOLOGY

How Elo Ratings & Improvability are Computed

Explore the mathematical formulations behind our scale-invariant TabArena Improvability regret metrics, stationary Bradley-Terry Elo tournament ratings, and 1,000-sample bootstrap confidence intervals.

LOCAL WEBGPU MACHINE LEARNING

Evaluate Tabular Foundation Models In Your Spreadsheets

Carla HQ brings TabICLv2 directly into Google Sheets via in-browser WebGPU inference. Run zero-trust predictions on private data without API keys or cloud server uploads.