Everything you should know before your first query.
Octavia is built around a simple promise: pursue the best-quality practical output while keeping the price as low as the work allows.
What is Octavia Works?
Octavia Works is a pay-as-you-go AI service designed to produce strong, useful answers without requiring a monthly subscription. It can analyze, research, explain, draft, compare and solve complex requests through a multi-stage processing pipeline.
How is Octavia different from a single-model chatbot?
A complex request does not always go to one model as one undivided prompt. Octavia can audit the request, separate it into focused parts, assign each part to the model best suited to handle it, and then synthesize the results into one coherent response.
Why can Octavia offer high-quality output at a lower price?
The system is engineered to reduce unnecessary work while protecting answer quality. Depending on the request, Octavia may use query decomposition, selective model routing, parallel work, bounded rework and eligible caching so that expensive processing is used only where it adds value.
Which AI models does Octavia use?
We do not publish a fixed model or vendor roster. The available systems and routing strategy can change as quality, capability, reliability and cost change. Keeping the roster flexible lets Octavia choose the strongest practical option for each portion of a request.
Does Octavia use the same model for every part of a query?
Not necessarily. One portion may need careful reasoning, another may need structured extraction, another may need concise drafting, and another may need synthesis. Octavia can route those portions differently and then combine the work.
What does it mean that a query is audited before processing?
Before work is dispatched, an automated audit examines the request for scope, structure, constraints and the type of work required. This helps determine whether the request should be answered directly or broken into multiple tasks, and helps reduce omissions and duplicated effort.
Do responses stream?
Yes. Once the final response begins, Octavia streams it progressively to the screen instead of waiting for the entire answer to finish before showing anything.
How does pay-as-you-go pricing work?
The first 10 eligible text-only queries for a new account are free. After that, Octavia shows an estimated price before you run a paid query. The amount is held from your wallet, actual usage is calculated after completion, and any difference is automatically reconciled.
Is there a subscription?
No. You add credit to your wallet and use it when needed. There is no recurring subscription fee for ordinary use.
Can I upload files?
Yes. A query can include up to three attached files. Supported formats are PDF (up to 20 MB and 75 pages), JPEG, PNG, GIF and WebP images (up to 5 MB each), Word documents (DOCX), Excel workbooks (XLSX), and text-based files such as TXT, Markdown, CSV and common source-code files (up to 10 MB each). Password-protected files, macro-enabled Office files, archives, executables, audio and video are not accepted. See the file-processing overview for more detail.
Do file queries use my free queries?
No. The free-query allowance applies only to eligible text-only queries. Queries containing uploaded files are not included in the free-query allowance and require an account with sufficient wallet credit.
How are file queries priced?
Before a file query runs, Octavia calculates a price on the server from the actual content being processed and shows you two numbers: the estimated charge and the maximum amount that will be temporarily reserved from your wallet. Only the maximum is held while the query runs; after completion you are charged for actual usage and the unused portion of the reservation is refunded automatically. File queries have a $0.10 minimum.
What happens to files I upload?
Uploaded files and the working materials derived from them (extracted text, document structure, workbook summaries and similar) are retained and may be used to develop and improve Octavia's AI systems, as described in the Privacy Policy and acknowledged before your first upload. Removing a file from the compose area before running a query only detaches it from that query. Do not upload content you are not authorized to provide.
How accurate is the price estimate?
It is a pre-run estimate based on the size and expected structure of the work. The final cost can differ slightly because the actual processing path and output length are known only after the query runs. The wallet is automatically charged or refunded for the difference.
What does caching mean here?
Where appropriate and safe, reusable non-personal computation or eligible repeated work may be cached so it does not have to be purchased and processed again. Private prompts and account data are handled according to the Privacy Policy and are not exposed to other users.
Does Octavia sell user data or use it for advertising?
Octavia never sells or shares the content of your requests, files, conversations or responses, and that content is never sent to an analytics or advertising partner. Separately, two optional cookie categories are off until you switch them on: Analytics loads Google Analytics, configured without ad-personalization signals; Advertising and retargeting loads the Meta Pixel, which Meta and Octavia use for advertising measurement, audience creation and retargeting. Under some privacy laws, turning on the advertising category counts as sharing personal information for cross-context behavioural advertising. You can change or withdraw either choice at any time from Cookie choices in the footer. See the Privacy Policy for the complete data-handling description.
Can I rely on an answer as legal, medical or financial advice?
No. AI-generated responses can be incomplete or incorrect. Important decisions should be independently checked and, where appropriate, reviewed by a qualified professional.