On TokenHub, different models may have different billing methods. The following introduces the billing methods for each type of model. You can also go to the model details page from Model Gallery to view the billing rules for each model. Language Models
|
Input tokens | Pay-as-you-go | USD / million tokens | Token consumption of user input text (including system prompt) |
Output tokens | Pay-as-you-go | USD / million tokens | Token consumption of model-generated text |
Cached Input | Pay-as-you-go | USD / million tokens | Cached input Token consumption |
Billing Notes:
When an inference request is interrupted (including active interruption by the user or passive interruption caused by other reasons), Tokens that have been generated by the server but not fully output are also billed.
Some models support tiered pricing (such as different unit prices for different input length ranges).
Some reasoning models' thinking process (Reasoning) and regular output are priced separately.
Prerequisites
Each time you claim a free resource package for a model or enable a model's postpaid service, a portion of your account balance will be automatically frozen (USD 30 for Hunyuan language models and USD 10 for other language models). If your account balance or credit limit is insufficient, the claim or enablement will fail.
The frozen amount will be refunded to your account after the resource is terminated. To refund the pre-frozen amount, delete all endpoints of the model (including default and custom inference services) on the Online Inference page. Note that after endpoints are deleted, the related model services will become unavailable. Proceed with caution. References
For detailed pricing information of each model, see Model Pricing. For details on the impact of billing arrears and recovery mechanisms, see Overdue Payments.