Understand model pricing
Where to find prices
Sign in to the model marketplace, select the same group as your API key, and find the model.
Prices, available models and plans can change. Refer to the current marketplace, payment confirmation and actual usage records.
Fields to check
| Field | Meaning |
|---|---|
| Input price | Unit price for content sent to the model |
| Output price | Unit price for generated content |
| Unit | Check whether prices are per 1K or 1M tokens, per request, per image, etc. |
| Cache price | Separate rules may apply when the model supports caching and a cache is actually used |
| Group | The same model can have different prices in different groups |
Image, audio or tool usage may have additional billing items. Read the specific model's description.
Estimate a text request
Basic cost ≈ input tokens / pricing unit × input unit price
+ output tokens / pricing unit × output unit price
This is only a basic estimate. Caching, other input types, tools and group rules can affect the result. Check the request's usage record for the final cost.
Why input usage grows in a long conversation
Clients often send previous messages with each request. A short new message may also include system prompts, conversation history, attachments or retrieved knowledge.
Start a new conversation when old context is unnecessary. Adjust history and output limits where appropriate, and check automatic retry settings.
Avoid unexpected spending
- Start with short inputs and gradually increase usage.
- Give applications separate keys with spending limits.
- Watch automatic retries, background agent tasks and concurrent requests.
- Regularly review usage logs by key.