Our NLP SaaS platform operates on a transaction-based pricing model. Clients are billed based on their usage of a unit known as 'credits'.
Credits are calculated using a credit conversion table that considers various factors, such as the origin, the type of transaction, the context, the specific large language model used, and the amount of context chunks utilized in the transaction.
For textual documents such (pdf, doc, docx, ppt and pptx) one context chunk is one page of the document. For transactions involving video or audio content, the amount of context chunks is assessed based on the duration of the content injected as context in the transaction. In the case of video or audio files, one context chunk corresponds to a one-minute segment of the file content.
If the transaction has no context injection, the amount of context chunks is one.
Depending on the subscribed plan, you will receive a monthly credit allowance. Any transaction exceeding the monthly allowance will be charged at the credit rate specified in each plan outlined above.
For instance:
Let’s say that you are in our platform on the web (that is from where the transaction is started, `web`) in a chat session (that’s the context, `chat`) with a document that has 50 pages, and you ask for something (that’s the transaction type, an `inference`) of the first 3 pages (that’s the `context chunks`) of such document using the model `gpt-4o-mini`: The total amount of credits used in this transaction will be 1 (the base credit rate for this transaction) times the number of context chunks that is 3 for this example, therefore the final amount of credits used will be 3.
| Started from | Transaction type | Transaction context | Model | Cached | Factor |
|---|
Yes, you can. When you subscribe to a plan with a monthly allowance included, you will have the option to set whether you allow usage beyond the contracted plan or not.
We do not permit usage beyond the monthly allowance by default, but you can change this setting at any time in your account settings.
No, cached transactions are not charged. When a transaction is cached, it means that the data and results of that transaction are temporarily stored on our platform for quicker access in the future. You can access the cached transaction data within the cache retention period (typically seven days) without incurring any additional charges. This feature allows you to optimize your usage and minimize expenses by avoiding duplicate processing of the same data.
The cache retention period refers to the duration for which we store cached transactions on our platform. Typically, we retain cached transactions for seven days. During this period, you can access and leverage the cached information as needed without any extra costs. This not only enhances the speed and responsiveness of your NLP applications but also helps you save on transaction charges by reusing the cached data when applicable.