Images & video
Send images to vision models through chat completions or messages. Image generation, video and audio are not available yet.
La documentazione è in inglese.
What works today, and what doesn't yet.
Image input (vision)
Models with the vision category read images. Send them as content parts in /v1/chat/completions, or as image blocks in /v1/messages, the same way as with the OpenAI and Anthropic APIs. Use a public HTTPS URL or base64 data. The model provider counts the image as input tokens, and we bill those like any other input token.
reply = client.chat.completions.create( model="openai/gpt-6-astra", messages=[{ "role": "user", "content": [ {"type": "text", "text": "What is in this picture?"}, {"type": "image_url", "image_url": {"url": "https://example.com/photo.jpg"}}, ], }],)print(reply.choices[0].message.content)A request body can be up to 32 MB, base64 data included. For large images, a URL keeps the request small.
Image generation
Not available yet. There is no /v1/images endpoint; calls to it return 404 not_found.
Video
Not available yet.
Audio
Speech, transcription and realtime audio are not available yet either.
New endpoints are announced in the changelog.