# Rerank API

> Cohere-style reranking with the MiniLM-L6 cross-encoder on Mojo CPU kernels, $10 per million pairs

MCP: https://rerank.mojojojo.cc/mcp (JSON-RPC over POST). OpenAPI: https://rerank.mojojojo.cc/openapi.json.
Auth: `Authorization: Bearer mj_live_...`; create a key at https://mojojojo.cc/keys. Usage is billed from mojojojo credits.

## Pricing

- Query-document pairs: $0.01 per 1k pairs (1 free per month)

## Endpoints

- `POST /v1/rerank` (MCP tool `rerank_documents`): Score documents against a query with a MiniLM cross-encoder and return them best first (higher logit is more relevant).

## Limits and measured results

- Scores are raw cross-encoder logits (higher is more relevant); not comparable across queries.
- Up to 512 documents per request; each document is truncated to 512 tokens with the query.
- Max logit difference to PyTorch 2.9e-06.
- Speed: 0.34x PyTorch at batch 1, 0.89x at batch 200. English MS MARCO passages only; evaluate on your own data.