Models

DocsRouter exposes OpenRouter models that currently declare image input, text output, and valid input/output pricing. The routing policies use a smaller recommended set, while explicit model selection can use any entry returned by the live catalog.

List models

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

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

The response uses an OpenAI-style model list and adds a docsrouter object with OCR-oriented metadata. Availability, names, context limits, and prices refresh from OpenRouter. Models that are text-only, unavailable, or missing valid pricing are excluded.

Routing model reference

These descriptions are routing guidance, not measured rankings or accuracy guarantees. DocsRouter does not publish an OCR benchmark dataset or ELO leaderboard.

Select an explicit model

Pass a supported ID to /v1/ocr:

{
  "url": "https://example.com/document.png",
  "model": "openai/gpt-4o"
}

An explicit model takes precedence over strategy. IDs that are absent from the current priced image catalog return a validation error.

Use a routing policy

If you omit model, DocsRouter uses strategy (default: balanced) to make a transparent, static selection:

{
  "url": "https://example.com/document.png",
  "strategy": "cost"
}

The policy does not classify document difficulty, compare outputs, retry failed extractions with another model, or perform a confidence-based waterfall.

The equivalent aliases docsrouter/quality, docsrouter/accuracy, docsrouter/speed, docsrouter/cost, and docsrouter/balanced are accepted as model IDs.

Pricing

Model-list prices include DocsRouter's 5% platform fee over the provider rate. OCR responses break billing into:

• provider_cost_cents
• platform_fee_cents
• total_cost_cents

The provider's upstream request cost is used when available. Otherwise DocsRouter calculates the cost from separate input and output token rates. Minimum charges and rounding apply.

Mistral OCR is a separate dedicated PDF-processing option. Its direct provider rate is $2 per 1,000 pages before the platform fee.

Other model endpoints

• GET /v1/models/{model_id} returns one supported model.
• GET /v1/models/stats returns aggregate data for the current live list.
• GET /v1/models/recommended returns models marked as recommended.
• GET /v1/models/categories groups the current list by category.