Skip to main content
Duomi AI uses a prepaid balance. After you add funds, each API request automatically deducts the amount calculated from the selected model, input tokens, output tokens, cached input tokens, and the applicable group multiplier.

​
How to add funds

  1. Open the wallet page and sign in to your Duomi AI account.
  2. Choose an amount in the Add funds section, or enter a custom amount. Common amounts include RMB 1, 10, 50, 100, 500, 1,000, 2,000, and 5,000.
  3. Select a payment method and complete payment. The wallet currently shows Alipay; use the payment options displayed on the live page. A RMB 1 top-up provides USD 1 of model quota calculated at official list prices.
  4. After payment completes, the funds are added to your wallet. You can review your balance and order history on the wallet page.
Top-ups start at RMB 1 and balances remain valid long term. They do not expire automatically over time.

​
Model pricing

Input, output, and cached-token prices follow the corresponding official model prices. The live pricing page is the source of truth because upstream prices and model availability can change. Under a unified exchange rate, Duomi AI applies different group multipliers based on the model and Coding plan. The same official usage can deduct different RMB amounts when the group multiplier is different.

​
Billing formula

The basic calculation is:
Where:
  • Official usage is the USD usage calculated from the model’s official input, output, and cached-token prices.
  • Group multiplier is determined by the model and Coding plan and represents the applicable discount or markup.
  • Balance deducted is the RMB amount charged from your Duomi AI wallet.
Using an estimate of USD 1 = RMB 7, before the group multiplier is applied, Duomi AI’s base charge is approximately 15% of the equivalent official RMB cost. The live model page and the actual group multiplier determine the final price. For example, if a request has USD 1 of official usage:

​
Cache-hit billing

When the upstream provider prices cache-hit input tokens at 10% of the regular input-token price, Duomi AI uses the same cache price and then applies the relevant group multiplier. Cache hits generally depend on whether reusable context prefixes remain identical across requests. The upstream model and protocol determine whether a request actually hits cache. Only the cache-hit portion uses the cache price; other input and output tokens use their respective regular prices.
For Coding workflows, we recommend using the official Codex or Claude Code client/CLI. These clients generally reuse conversation context and stable prompt prefixes more consistently, which can improve cache-hit rates and reduce the amount charged for the same work.

​
Billing example

Assume one request has the following official usage:
  • Non-cached input usage: USD 0.20
  • Cache-hit input usage: USD 0.50
  • Output usage: USD 0.30
  • Group multiplier: 0.3x
If the USD 0.50 cache-hit amount has already been calculated using the 10% cache price, the balance deduction is:
Actual requests can use different models, token counts, and group multipliers. Use the live model page, usage details returned by the API, and your wallet balance as the final reference.