Why am I being charged when I'm not making API calls?

If you're seeing unexpected charges on your account, you may be incurring costs from deployed models or endpoints that continue to run even when you're not actively using them. Understanding the different billing scenarios can help you identify and resolve these charges.

Common Billing Scenarios

There are three main ways you can incur charges, even when not actively using models:

1. Fine-tuned Model Hosting

When you deploy a fine-tuned model, you're charged per-minute hosting fees regardless of whether you're making requests to the model. These hosting costs are separate from the initial fine-tuning job costs and continue to accrue until you explicitly stop the endpoint.

2. Dedicated Endpoints

Dedicated endpoints are charged based on hardware allocation, even when there are no active requests or processing happening. The charges continue as long as the endpoint remains active.

3. Serverless Model Usage

Serverless models are charged based on actual API usage - specifically the tokens used and model size. This is the most straightforward billing method as you only pay for what you use.

How to Check and Manage Your Deployments

To avoid unexpected charges, regularly review and manage your active deployments:

1. Visit your models dashboard

2. Check for any deployed fine-tuned models

3. Look for active dedicated endpoints

4. Stop any endpoints you're not actively using

Monitoring Your Usage and Costs

You can monitor your usage and review pricing details at https://www.together.ai/pricing. This will help you understand your current billing and make informed decisions about your deployments.

By regularly managing your deployments and understanding these billing scenarios, you can avoid unexpected charges and maintain better control over your account costs.