Universal Pdf Support

OpenRouter says its PDF support works with any model on its service because it handles the document step itself. Per its April 2025 announcement, it automatically uses a provider's native PDF support where one exists (it names Gemini, Anthropic and OpenAI) and offers two processing engines. One is a free text-only extractor. The other, the default and priced per page, uses optical character recognition (reading text from page images), so it also handles image-only scans and extracts embedded images, which can then be sent to models that accept visual input. A developer therefore sends the same kind of PDF request whatever the model, while the engine chosen and the model's own input abilities determine what content the model actually receives.