Products that use more than one model

• Switching or mixing models without a new integration each time.
• One place to see spend per model, key and project.
• Predictable behavior when a model's service has an incident.
• Prices that can be checked against the labs' own lists.

• One OpenAI-compatible endpoint; the model parameter picks the model, and GET /v1/models lists the ids your key can call.
• Every model's per-token price is public, and leading models are priced at or below their official list prices.
• Usage can be broken down by project, API key and model, and exported as CSV.
• When a route fails, the request is retried on another route of the same model. It is never switched to a different model.
• Request fields that are not applied are named in the x-boostrail-degraded response header instead of being dropped silently.

1. Point your OpenAI SDK at https://api.boostrail.com/v1 with a BoostRail key.
2. Pick model ids from the Models page.
3. Use separate keys per environment or feature to see spend split the way you need.

• Chat Completions accepts n = 1 only.
• audio, web_search_options and non-text modalities return 400 unsupported_parameter.
• Non-streaming requests stop at 180 seconds; stream long generations.
• Some models carry an input cap, marked on the Models page.

Models and prices →Chat Completions reference →Errors and failover →All use cases →