FAQ: charges, balance and usage records
Common questions about charges, balance and usage records.
How am I charged?
Each request is charged by actual usage: input tokens and output tokens, each multiplied by the model’s price (USD per million tokens), deducted from your balance. See the Model pricing page for prices; usage records in the console are the record of what was charged.
Why does one message create several charge records?
Coding tools such as Claude Code, Codex and OpenCode often send several requests in the background after you send one message: reading files, running commands, calling tools, managing context, generating a conversation title, and so on. Each request is recorded and charged separately, so several records per message is normal.
My input was short. Why isn’t the cost low?
What you type is only a small part of each request. Coding tools include the system prompt, the earlier conversation and relevant file contents every time, so the input tokens actually sent are often far more than what you typed. The longer the conversation and the more files read, the larger each request’s input.
Ways to use less:
- Start a new conversation when a task is done, instead of continuing a long old one.
- Let the tool read only the files and directories it needs.
- Choose a cheaper model for simple tasks.
What are cache reads and cache writes?
Some models support prompt caching. The first time a long piece of context is sent, it is written to the cache; later requests that reuse it are billed as cache reads, usually much cheaper than normal input. Writing to the cache has its own price. Whether caching happens, and the exact price, are shown in the console’s usage records.
Why is my balance shown in USD?
Model prices are in USD, so the balance is shown in USD too. Top-ups are paid in CNY; the page shows the payment amount and the credited amount before you pay, and the order is final.
What happens when my balance runs out?
Requests return 403 with error code INSUFFICIENT_BALANCE. Top up and continue; you do not need a new key.
Where can I see my usage?
Log in to the console. Usage records show the time, model, token counts and cost of each request.
When should I contact support?
If usage records and charges do not match, or you see a charge the pages cannot explain, contact support using the details at the bottom of the page, with the request time and model name.
QQ user group: 1109931264