Two units, two ways to read usage
Your plan can be expressed in credits or tokens. Check the unit beside your balance before starting a request. A credit is not a token, and there is no fixed conversion that applies to every action.
Tokens measure an amount of content processed by AI. Credits are a product usage unit. The number of words in a reply alone therefore cannot tell you the credit charge. Identify your plan’s unit first, then check the usage recorded in your account.
Current plan allowances
The table below shows active public offers and their currently published allowances. It refreshes when you open this page and every 30 seconds while the tab remains visible. Use Refresh after a plan change to check the values immediately. The timestamp indicates the last successful retrieval.
These values follow the offer settings saved in the administration console. They are not your personal balance: an account may have individual conditions, an older offer, reservations or a top-up. For your own situation, open Plan and credits.
A dash in the caps column means this table does not indicate an additional cap. It does not mean unlimited usage. If refreshing becomes unavailable, the last received values are marked as potentially outdated.
How tokens are counted
On a token allowance, consumption includes input tokens and output tokens. Input is the content sent to handle the request: your message and relevant context, which may include part of the conversation, a strategy, code or an image. Output is the content generated and counted for that request.
A token is not necessarily a word: language, numbers, punctuation and code affect how content is divided. Your latest typed message may not represent the entire input. The visible answer may not represent the entire counted output either, particularly when the model uses reasoning.
Illustrative example, not a price quotation:
Input tokens: 1,200
Output tokens: 300
Usage: 1,200 + 300 = 1,500 tokensThis example explains the addition; it does not predict a prompt’s cost. Check recorded usage for the amount associated with your own request.
How credits are consumed
On a credit allowance, the charge depends on the AI task and the applicable consumption rules. A short explanation, an image interpretation and generation of a complete code file can use different amounts. The model used and the quantity of content processed can also affect the charge. A minimum consumption may apply to a request.
Publishing a universal number of credits per message or line of code would be misleading. Compare similar uses in your history, then adjust your workflow. The table’s allowances are budgets available per period, not a fixed price per answer.
| Your request | What helps keep usage focused |
|---|---|
| Explain a rule | Quote the rule and ask a precise question |
| Modify a strategy | Name the change and what must stay the same |
| Analyse a screenshot | Crop to the relevant area and ask one main question |
| Generate or fix code | Identify the version and provide the relevant full error |
| Understand a report | Choose the metrics and the decision you want to prepare |
Read your balance and history
In Studio or chat, open the credit or token counter. The usage panel shows the available amount, period allowance and, where applicable, caps, carry-over and top-ups. It may also show availability for the selected model.
To examine recorded use, open usage history. Compare the date, action and unit. To check your offer, visit Plan and credits. Saving a conversation does not mean each message has the same charge; switching models can change usage.
Understand a pending reservation
A task in progress may display a reserved amount. This temporarily reduces availability so several operations do not commit the same budget. Do not add the reservation a second time to the final charge for the same operation.
Once processing ends, read the updated status and recorded usage. An interrupted reply on screen alone does not establish that no processing occurred or that a final charge was made. If an amount appears to remain reserved after the task ends, refresh the panel and report the discrepancy with the time and displayed identifier, rather than sending many identical requests.
Allowances, caps and availability
The allowance is your budget for a period. Additional caps may distribute usage over a day, five hours or seven days. These caps are included within the allowance: they are not extra budget. A restriction for a model or model category may also limit availability shown in your account.
Illustrative example: a monthly allowance of 10,000 units with a five-hour cap of 1,000 units does not give you 11,000 units. When the five-hour cap is reached, an action may need to wait even if monthly budget remains. Rolling windows refill progressively; use the next refill time shown in your panel.
Renewal, carry-over and top-ups
A daily allowance renews at 00:00 UTC. A monthly allowance renews at the start of the calendar month, at 00:00 UTC. A one-time allowance has no automatic renewal. Your computer’s time zone may display a different local time. Allowance renewal can differ from your subscription billing date.
When carry-over is enabled on a monthly offer, unused units may be kept within the stated cap and validity period. Without applicable carry-over, do not assume unused budget accumulates. Purchased top-ups appear separately where available; check their unit, price and conditions before buying. Credits and tokens are not interchangeable by default.
When availability is insufficient
First identify the restriction: an exhausted allowance, a temporary cap, feature access or a model restriction. Upgrade / View plans opens the available options; it does not mean every restriction requires a purchase. A temporary cap may become available at the next indicated refill.
Write focused requests to manage your budget
Provide the useful context once, then request a targeted change. Avoid attaching a large unchanged file again when the conversation already provides sufficient context. A shorter answer may reduce some generated content, but no special word guarantees a fixed charge or a particular saving.
For the current version, explain only the exit rule.
Do not modify the strategy or generate code.
Give one buy example and one sell example, then identify
ambiguities to clarify before testing.Change only the EMA period from 20 to 30.
Keep the entries, protections and other parameters unchanged.
Summarise the changes before I rerun the backtest.An analysis requested from Copilot is an AI use distinct from reading an existing historical report. For precise requests, continue with Work with Copilot; to interpret results, see Understand metrics.