§ Pricing

Pay per verified label.

No setup. No minimums. No vendor calls.

§ 01 — Per-task rates

Per label. Per Human. Per task type.

Pricing under review
Per-task rates land before public launch. The structure below is final; the numbers are provisional.
Task typePer label (Builder)Per label (Human)Network fee
Image bboxTBDTBDTBD
PolygonTBDTBDTBD
SegmentationTBDTBDTBD
KeypointTBDTBDTBD
Text NERTBDTBDTBD
Text classificationTBDTBDTBD
SentimentTBDTBDTBD
RLHF preferenceTBDTBDTBD
WritingTBDTBDTBD
Audio transcriptionTBDTBDTBD
Audio classificationTBDTBDTBD
Video frameTBDTBDTBD
Video segmentationTBDTBDTBD
3D point cloudTBDTBDTBD
Document parsingTBDTBDTBD
CoreferenceTBDTBDTBD
Relations / moderationTBDTBDTBD
§ 02 — Consensus

3 Humans by default. Dial it where you need it.

1

Cheap-draft

1 Human, no peer check. Internal exploration, low-stakes prototyping.

3

Default

3 Humans, majority verifies. Balanced cost and confidence. Most workloads.

5

High-stakes

5 Humans, supermajority verifies. Safety, medical, legal, fairness-sensitive work.

7

Max confidence

7 Humans. Reserved for evals, gold-standard sets, regulated benchmarks.

Read how consensus works →
§ 03 — Network fee

What the network keeps.

The Network takes a transparent percentage of every Builder spend to fund operations, security, anti-gaming, payment rails, and the public Roadmap. The remainder goes directly to the Humans who verified the labels. Final percentage published before launch.

§ 04 — Inclusions

What every dollar buys.

  • Per-label rate paid to verified Humans.
  • N-Human consensus — every label checked by your chosen N.
  • Steward adjudication on disagreement.
  • Gold-question calibration baked into every batch.
  • Signed receipts and full audit trail.
  • Dataset export in 16+ formats.
  • Direct payouts via Wise, Razorpay, Rapyd, Stripe Connect, fiat to USDC.

Try it with 100 labels.