Credits
Credits are the unit of AI consumption in Hubrix. Every time an AI model processes a request — in Smart Chat, an agent response, a workflow prompt step, a research report, or a bulk job row — credits are consumed.
What consumes credits
| Feature | When credits are consumed |
|---|---|
| Smart Chat | Per AI response generated |
| Custom Agents | Per agent response generated |
| Workflows | Per Prompt, Agent Call, or AI-powered step that runs |
| Deep Research | Per completed report (varies by depth level) |
| Bulk Processing | Per row processed |
Operations that do not consume credits: browsing the UI, viewing documents, reading past reports or runs, commenting, or using the folder picker.
Typical credit costs
Most features have a per-operation credit cost — for example a Smart Chat response, a translation, a document (RAG) query, an agent run, a research report (standard or deep), a bulk-processing batch, or voice (text-to-speech and speech-to-text). Lighter operations cost the least; heavier ones such as agent runs and deep-research reports cost more.
The exact per-operation credit costs are configured per workspace and can change, so this page intentionally does not hard-code them. The current cost of an operation is shown as an estimate in-app before you run it (for example, the Deep Research dialog shows an estimated range), and the authoritative figures live in Settings → Billing. Token-heavy features such as chat and agent runs also scale with the AI model you choose and the length of the request — see below.
How credit costs are calculated
Credit costs are driven primarily by the number of AI tokens processed:
- Input tokens — everything sent to the AI model (your message, system prompt, retrieved document chunks, conversation history)
- Output tokens — the AI's response
Different models have different costs per token. Generally: more capable models (GPT-4o, Claude Opus) cost more credits per token than smaller models (Mistral, Deepseek).
The credit cost of an operation is shown as an estimate before you run it where possible. For example:
- The Deep Research setup dialog shows an estimated range based on the selected depth level and sources.
- The Bulk Processing wizard shows an estimated cost per row after you configure the operation.
Credit estimates are approximate. Actual cost may vary slightly because the exact number of tokens depends on the content of the AI's response, which cannot be predicted in advance.
Credits are deducted on completion
Credits are deducted after an operation completes, not when it starts. If an operation fails, you are not charged for work that was not done.
For bulk jobs: credits are deducted row by row as each row completes. Cancelling a job mid-run means you only pay for the rows that finished.
Viewing your balance
Your current credit balance is shown in:
- Settings → Billing (detailed view with usage history)
- The top bar of the Hubrix UI (small indicator when balance is low)
Topping up credits
Admins can top up credits in Settings → Billing → Add credits. Hubrix offers several standard top-up packages of increasing size, each with its own credit amount and price. The available packages and their current prices are shown in the Add credits dialog.
Top-up package sizes and prices can change, so they are not hard-coded here — the current options are always shown live in Settings → Billing → Add credits.
Purchased top-up credits are added immediately to the workspace balance and are available to all members. Unlike monthly plan credits, top-up credits carry over and do not expire at the end of the billing cycle.
Top-up credits do not expire on the monthly reset date — they persist until used. Plan-included credits reset monthly and do not roll over; top-up credits are separate and carry over.
Was this helpful?