← Back to Blog

Archived Concept - Future OCR Models and Stable Integrations

by DocsRouter Team

Historical and speculative

This post previously speculated about an unreleased model. It was not based on a DocsRouter benchmark and should not be read as product documentation or an availability announcement.

The durable lesson is simple: keep model choice configurable and discover supported IDs from the API instead of hard-coding an announced or rumored model name.

curl https://api.docsrouter.com/v1/models

DocsRouter exposes the currently priced image-input models returned by OpenRouter. A release becomes usable only when it appears in GET /v1/models and is accepted by /v1/ocr; being ranked on an external leaderboard is not enough.

The static quality, accuracy, speed, cost, and balanced policies can reduce application configuration, but they do not provide automatic benchmarking, result comparison, or fallback. Evaluate available models with your own documents before changing a production model selection.