or-find-vision-models

name	or-find-vision-models
description	Use when the user wants to find vision-capable / multimodal image-input models on OpenRouter. Triggers on phrases like "OpenRouter vision models", "image-capable models on OR", "models that accept images on OpenRouter", "multimodal models on OpenRouter", "OCR-capable OR models", "find a vision model on OpenRouter for <task>".

Find OpenRouter Vision Models

Filter the OpenRouter catalog to models that accept image input, then rank or summarize based on user criteria.

When to use

The user wants to discover models that can process images on OpenRouter — for OCR, image understanding, visual QA, document parsing, screenshot analysis, etc.

Workflow

1. Fetch the catalog:

   curl -s https://openrouter.ai/api/v1/models -H "Accept: application/json"
   

2. Filter data[] where architecture.input_modalities includes "image".
3. Note the pricing.image field — many vision models charge a separate per-image fee on top of token costs. Surface this in output.
4. If the user expressed preferences (cheapest, largest context, specific use case), apply ranking:
   • Cheapest — sort by combined pricing.prompt + pricing.image
   • Largest context — sort by context_length descending
   • Best for OCR/documents — favor models with high context windows and known strong OCR (Gemini, Qwen-VL family); flag this from the model description
5. If criteria are vague, ask one short question: budget? volume of images? text length per request?
6. Return top 5–10 as a markdown table: ID, Context, Prompt $/1M, Completion $/1M, Image $ each, Other input modalities.

Notes

• Some models accept images but charge per image (not per token). Always show both costs.
• A model listing image in input but not output is for image understanding only — not generation. Image generation models also list image in output_modalities. If the user wants generation, route to that subset.
• Flag if a model's vision capability is limited (low max image size, or supports only small batches) when this info is in the description.

Output

Markdown table sorted by the user's criterion (default: cheapest combined cost). Include a count of total vision-capable models in the catalog.