Model selection

Choose the API family or call entry first, then choose the model product:

• OpenAI Chat Completions: suitable for many existing OpenAI-compatible clients.
• OpenAI Responses: suitable for native Responses tools and agentic workflows.
• Anthropic Messages: suitable for Claude Code and Anthropic-compatible clients.
• Gemini GenerateContent: suitable for native Gemini request structures.
• Images Generations: for image generation through the public API.

Image-to-image openai_images_edits is not a public API option. It can only be called by the signed-in web page through POST /api/img/edits; do not integrate it as a public model API or as the OpenAI SDK Images Edits endpoint.

The model name must be the current public model id of a model product enabled for the call entry you are using. A model with the same name visible under another call entry is not proof that it is available here.

When choosing, consider quality, latency, input and output prices, the cache-read and cache-creation prices supported by the model product for the current call entry, image pricing, and upstream stability. Validate with a minimal request before increasing traffic.