Skip to main content
The Gateway also routes to frontier vision AI models via upstream providers, so you can use the same API key and baseURL to access them as well. These models are passthrough VLMs supporting the same methods and method_params reported by their providers. On GET /v1/openai/models a frontier model carries an upstream vendor in its provider field. Which vendor serves a given model can change as capacity and price move, so read it from the catalog at call time rather than pinning it here. Prices are the provider’s listed prices in USD per 1M tokens (input / output).

Prices in USD per 1M tokens

Every frontier model accepts a JSON Schema response_format, except meta/muse-glimmer-30b, which accepts json_object only. See Methods & Response Format. qwen/qwen3.8-27b is not listed here. It runs on VLM Run GPUs, so its provider reads vlm-run rather than an upstream vendor, and it is documented with the Gateway VQA models. Treat vlm-run, and an empty provider, as the two ways the catalog says “self-hosted”.

​
Next steps

Models

The full Gateway model catalog.

Chat Completions API

Full request parameters, streaming, and error handling.