Send images to vision models through the OpenRouter API (multimodal image inputs)

Documented steps

1. Use the content-array message form: {"role": "user", "content": [{"type": "text", "text": "<prompt>"}, {"type": "image_url", "image_url": {"url": "<https URL or data URL>"}}]} — docs recommend text first, then images.
2. For local files, base64-encode and pass a data URL: data:image/jpeg;base64,<encoded bytes> (also image/png, image/webp, image/gif).
3. Pick a vision-capable model: GET /api/v1/models and filter input_modalities for image, or check the model page.
4. POST to /api/v1/chat/completions as usual (Bearer auth); the model answers about the image in choices[0].message.content.
5. Image tokens are billed via the model's image pricing field (see GET /api/v1/models pricing.image).
6. Docs: https://openrouter.ai/docs/guides/overview/multimodal/image-understanding