AI model credit rates

Credits are the only unit you need to think about in the product. Intelligence tools run against your organization’s credit balance; when a run uses language, image, or video models, the platform measures tokens, images, or seconds and converts them into settled credit debits using the same catalog basis summarized below.

Use this page when you compare models, sanity-check a heavy job, or align finance with ops. Everyday tasks like topping up stay on Credits & billing.

Credits as the platform unit

• Balance lives on the organization, not your personal login.
• Subscriptions and top-ups increase that balance in whole credits.
• Each run either charges a predictable credit fee for the tool (see Tools overview) or settles against metered usage when you pick premium models or generate lots of media.
• Fractional usage still stacks up in credits—token tables below use per 10,000 tokens so numbers stay easy to relate to typical prompts.

Fixed tool fees vs metered settlement

Many tools advertise a flat credit cost per execution so planning stays simple. When you override models or produce long outputs, settlement follows actual tokens / images / seconds from your run against the same reference rates. A small authorization hold may apply when the run starts; the final debit lines up with metered usage and can be higher when usage is heavy—your run detail is the authoritative breakdown.

How to read the tables

All numeric cells are credits. Headers state the denominator: 10k tokens, one image / MP, or one second of video.

Text & token-priced image (per 10k tokens)

Three columns—input, output, and read (for repeated/cached context where applicable). Figures are credits per 10,000 tokens. If read is marked “—”, that model does not expose a separate read line in this reference.

Native image (per image or per megapixel)

The native block lists total credits per generated asset—either one image or one megapixel, as shown in the basis column—when billing is not token-based.

Video (per second)

Credits per second of generated output. Estimate total credits by multiplying rate × duration in seconds; some SKUs add tiers—see the footnote and your run summary when the catalog shows composite pricing.

Browse rates by modality

Switch tabs for text, image, or video. Image combines token-priced models (same per-10k columns as text) with native per-asset pricing.

What can add credits beyond these rows

Even with the same model, runs can cost more when any of the following are included:

• Tool or function invocations billed separately
• Web search or grounding add-ons
• Agent or deep research style jobs with extra steps
• Reasoning or batch surcharges on certain SKUs

Those lines roll into the same credit settlement as base tokens or seconds.

Overrides vs defaults

When the UI offers a model override, only approved options are selectable and validated. If you skip overrides, Pinnora chooses a default for that tool.

After your run — where totals appear

Completed runs surface usage and settlement in the workspace so owners can reconcile balances. For disputes or anomalies, capture the organization slug, project, tool ID, and timestamp before contacting support.

Freshness

Vendor catalogs change. We refresh this reference when our pricing snapshot moves; your billing and run views always reflect what actually settled for that execution.

Related articles

• Credits & billing — balance, top-ups, plans
• Runs & status — statuses and failure handling
• Tools overview — tool matrix and typical per-run credits