Seton Labs · Dataset tiers

Tiers are zones,
not fixed points.

Three axes define the space: how many rows a dataset has, how many categories and subcategories it covers, and how much compute it consumes per model answer. A tier is a region inside that space — two datasets in the same tier can still land at different coordinates.

Rotate the dataset space Drag to orbit · pinch or scroll to zoom

Each point is one example dataset. Tiers form loose clusters — same tier, different coordinates.

15 reference points

What each axis means

X — Rows

Total rows in the dataset across our organization.

Y — Categories

Total count of categories and subcategories covered.

Z — Compute cost

Extra tokens a model generates per answer for this dataset type.

Tier colors

Each tier occupies a different region of the space.

EffortlessSmallest workloads

EasyLight and manageable

MidMiddle complexity

HardHeavier datasets

UltraHardMost expensive region

Key idea

A tier is not fixed coordinates

Two datasets in the same tier differ in rows, categories, and compute cost. The tier is the neighborhood; the dataset is one address inside it.

Why this matters

Budgeting and routing decisions use the tier label, but actual cost and size vary inside that label. Knowing the range prevents surprises.

Keyboard: ←→↑↓ rotate · +− zoom · R reset

Tiers are zones,not fixed points.

Tiers are zones,
not fixed points.