How credits work

Credits meter usage: stronger models, larger datasets and deeper runs use more.

Credits meter usage. Stronger models, larger datasets and deeper runs use more credits. Most everyday actions use none.

What credits cover

These actions deduct credits as you use them:

  • Premium and reasoning models: every token going to a premium model deducts from credits. Reasoning models (the kind with explicit thinking-effort dials) consume thinking tokens too.
  • Chats over larger datasets: when the documents no longer fit in one pass, the model works in several steps, and each step uses credits whatever the model.
  • Workflow runs: Summarizer, Data Extractor, Blueprint, Humanizer, SEO Metadata, Individual. Usage scales with dataset size and configuration.
  • Enhanced OCR: 1 credit per page in the document, billed up-front when you trigger the workflow.
  • Compose renders for some formats: most artifact generation is included; very large or unusual renders may meter.

What credits don't cover

These don't touch credits; they're covered by your plan's flat allowance:

  • Document uploads.
  • Storage (within plan quota).
  • Automatic OCR.
  • Chats on the default models over a small set: they round to zero.
  • Document search and retrieval.
  • The document viewer, the info panel, and label/list operations.
  • The API call layer itself.

If you hit a plan limit (uploads per month, chats per day, etc.), it's not a credit problem; see Plans and what credits buy you.

Monthly allowance, plus bundles

Two ways credits enter your account:

  • Monthly allowance: your plan includes a monthly credit budget. It resets on the first day of each calendar month, and unused credits do not roll over.
  • One-time bundles: buy a bundle on top, sizes vary. Bundles don't expire: they stay in your account until used.

Allowance is consumed first; bundles fill in when allowance runs out. You'll see a clear breakdown in the credit dashboard.

What happens when credits run out

Depends on your plan:

  • Free, Basic: premium-model and workflow actions stop until next month's allowance. You can buy a bundle to keep going.
  • Pro, Team, Enterprise: overage handling can be configured (allow / cap / require approval) in billing settings.

You'll see warnings as you approach the limit; nothing is silently cut off.

Predictability

Workflows and Enhanced OCR show the estimated credits before you confirm ("this Summarizer run will use ~24 credits"). The estimate is accurate to within a few credits; actual usage depends on document size and content density.

For premium-model chats, typical credits per turn show in the model picker so you can pick deliberately.

What's next

Was this helpful?