AI quotas
How the token quota works, what consumes it, and what happens when it runs out.
How quotas work
AI tokens govern the use of language-model-backed features. The quota belongs to the whole organization — every user draws from the same pool — and it resets at the start of each billing cycle. There is no rollover: unused tokens do not carry into the next cycle.
Quota per plan
Plan | Tokens per cycle |
|---|---|
Essencial | AI not included |
Profissional | 7,500,000 |
Enterprise | 18,000,000 |
Degustação | 500,000 |
The Degustação quota is deliberately far smaller than Profissional's. It exists to take a full tour of the product, not to run a portfolio — every other Degustação limit mirrors Enterprise.
What consumes tokens
Validating a PARTIAL or NA answer: a few hundred tokens per validation, handled by a fast model. Generating a report: on the order of thousands to tens of thousands per report depending on complexity — a P3 supplier, answering 257 controls, produces far more content than a P1 answering 79. Conversations with Ara: they vary with how many tools the answer has to call.
Aranis does not use a single model. Calls are routed across tiers — a fast one for short, well-scoped tasks and a reasoning one for analysis and long-form writing — and the model behind each tier is set at the platform level, so it can change without altering your plan or your quota. Cumulative consumption for the cycle is visible under Settings › Subscription.
When the quota runs out
Once the limit is reached, AI operations pause until the next cycle or until a plan upgrade. In practice: PARTIAL/NA validations stay pending, report generation is blocked, and Ara reports that the quota is exhausted. Everything that does not depend on AI keeps working normally — browsing, reading data, answering YES and NO, exporting.