Bring your own model
Bring your own model means an application's AI steps call the model provider you choose, with your credentials, and the usage is billed directly to your own provider account rather than resold as part of the application's fee.
Why it matters
When a platform bundles model usage into its own price, you pay its margin on every call, your usage is capped or metered in its units, and the choice of model is its choice. As document volume grows, so does the gap between what the platform charges and what the provider would.
Bringing your own model keeps the provider relationship yours: your rates, your data-processing terms with that provider, and your choice of model.
How Runtara applies it
On every plan, including Runtara-managed Cloud, model calls in Runtara use your credentials and your chosen endpoint, and they are billed directly to your provider account.
- Your provider, your account. Supported connections include OpenAI and AWS Bedrock; a customer-managed or air-gapped deployment can use a model endpoint inside your environment.
- Separate from the Runtara licence. Model usage stays on your provider bill. The Runtara subscription does not include or mark up model costs.
- Change provider through configuration. Switching the model behind a step is a configuration change, not a rebuild.
- Cost attributed to each run. You can see which model calls contributed to each run’s cost. In an illustrative order run, three calls total $0.05818 at the example provider rates; actual cost depends on your models and usage.
What the model can do
Bringing your own model does not widen what the model is allowed to do. Each AI step chooses only from the tools compiled into the workflow when it was authored. The security model explains how that boundary works.