Account and billing

Settings → Account & security holds your sign-in email and password. Settings → Usage & billing holds your plan, monthly credits, top-ups and payments, with what you have used this month. Sign in

Everything that calls a model draws from the same credits. Chats, Galaxy Lab, questions about a review and literature reviews are charged by what their model calls cost; the co-author, Studio outputs, the planning and reviews of visual summaries and AI site-design requests by their tokens, 1 credit for each 1,000 tokens (the text a model reads and writes). Your balance is in Settings → Usage & billing.

A chat turn or a literature review (its figure readings included) is charged by what its model calls cost instead of by their tokens. Each call is priced at its model’s list price, separately for the text it reads, the text it rereads from the model’s cache and the text it writes, reasoning included. The total is converted at 1 credit for each US$0.000275, the cost of 1,000 tokens of GLM 5.3 Flash, 800 read and 200 written. Text reread from the cache costs a fifth of new text on GLM 5.3 Flash and a tenth on GPT-6 Luna, so a long chat, which rereads its earlier steps, uses fewer credits than its tokens; a review that reasons and writes at length uses more. On GPT-6 Luna, new text that the model keeps in its cache for later steps costs a quarter more than other new text. The line under each answer shows what it was charged.

An AI site-design request also pays for the time the designer works on it: 9 credits a minute, counted by the second, on top of its tokens. A request of about a minute is about 9 credits more; there is no monthly number of requests.

Credits also buy one more Studio presentation or Visual summary once the month’s are used up. The notice shows its price and your balance, and Use N credits starts it (for a visual summary, Hold N credits, since it is paid by the picture); with too few credits it offers Buy credits and See plans instead. The price is held first and charged only when the output is ready, so a job that fails gives it back. Its model tokens are charged as credits too, as for everything else. Each price is on the Pricing cards.

Past the month’s code sandbox time, credits can keep your code running too: turn on Keep computing on credits in Settings → Usage & billing, and each minute a sandbox runs past the hours is charged at the price in the plans table, your monthly credits first and then top-up credits (Code sandbox limits).

Everything is paid in the same credits. A chat turn and a literature review cost what their calls cost the model, and the other AI calls their tokens; a few things also cost time or a picture:

Within your plan’s allowances, code sandbox time costs no credits, and presentations and drawings cost only their tokens. The plans table below shows each plan’s allowances and these prices as RBAi applies them.

Each plan’s price, monthly credits and limits, as RBAi applies them:

Every plan can use every part of the app. A count marked “a month” renews with the calendar month, India time. Code sandbox time and emails sent through RBAi are listed while they are limited; those emails are the requests RBAi sends for you to researchers outside RBAi (Connect). The RBAi Scholar and MCP Cloud APIs are described in API and MCP Cloud.

Your own plan, its limits and what you have used are in Settings → Usage & billing. The Pricing page shows the same numbers.

What File storage counts

File storage counts your uploads, the files answers make (plots, pictures and saved results), library PDFs, project source files and HTML pages. The disks of cloud computers and chat sandboxes (about 26 GB each) and memory folders (up to 5 MB each) are separate and do not count. Text does not count either: chats, reviews, notes, and the Word and Excel documents RBAi writes. Lookup results RBAi saves with your chats don’t count.

When your storage is full, answers still save their plots and pictures, so an answer never loses them; uploads and other new files are refused until you free space. The chat and Settings → Files say when it is full, and Settings → Files lists what uses it, largest first.

Each action is charged what it uses. A chat question, a Galaxy Lab request, a question about a review and a literature review are charged by what their model calls cost; the other parts use 1 credit per 1,000 tokens. So the same kind of request costs more when it is long or complex, and less when it is short.

Top-up credits never expire, and they are used only after your monthly credits. Buy them with Buy credits in Settings → Usage & billing, on any plan.

Paid plans renew monthly through Razorpay, by UPI AutoPay or card. Each billing period starts with a full month of credits, and unused monthly credits don’t carry over. Free credits renew on the 1st of each month, India time.

Top-up credits are used next. With none left, new AI requests pause until your credits renew, and the app offers an upgrade or a top-up. A literature review needs about 150 credits to start, and it can use your last monthly credits and your top-up credits together.

Work that is already running stops when the credits run out, and keeps what it finished: a chat turn ends between two steps with Stopped: your credits ran out, and a co-author task, a literature review and a presentation the AI is building stop with what they wrote. Each shows the same offer: Buy credits, and Upgrade on the Free plan. After a top-up, reply continue in the chat, or ask again.

You can move from Free to any plan at any time. Switching between paid plans in the middle of a period isn’t available yet. To switch, cancel in Settings → Usage & billing (your plan stays active until the end of the period you paid for), then choose the new plan. Top-ups add credits at any time, on any plan.

A cancellation and refund request made within 5 days of a purchase is considered, and an approved refund is credited to your original payment method within 5 days. Top-up credits you have started using aren’t refundable. The Cancellation and Refund Policy has the full terms.