Skip to main content
Every chat model in the catalog works on all three chat endpoints. Pick the format your code or tool already speaks, not the one the model’s vendor uses. A Claude model answers on Chat Completions, and a GPT model answers on Messages. Images, video, audio, embeddings and moderation have their own endpoints.

Chat formats

Each chat endpoint takes the format of one official API and works with its SDK: For new code, the formats differ in a few ways that matter: Responses is OpenAI’s current format, and the examples in these docs open on it. Chat Completions is the older format, for code and tools that can’t use Responses or Messages. Web search works only in Responses and Messages, see Web search.

Other endpoints

Images, audio, embeddings and moderation follow the OpenAI format, so the OpenAI SDK calls them unchanged. Video jobs follow the OpenRouter Videos API: Field-by-field contracts for every endpoint are in the API reference.