Cohere Rerank API Errors and Truncated Documents

Share
Workaround available API Updated November 14, 2026Based on 3 user reportsHow we research

Quick answer

Rerank errors are usually malformed requests (400/422), too large payloads, or documents cut at 4096 tokens. Keep requests under about 1,000 documents and chunk long documents.

Most likely fix: Validate the request

How to fix it

Try these in order — the most common fix is first.

  1. 1. Validate the request

    1. Send model, query and documents as the reference describes
    2. Format structured data as YAML strings
    3. Read the id and message fields in the error body
    Did this fix it?
  2. 2. Batch and chunk

    1. Split document sets above 1,000 into batches and merge scores
    2. Chunk long documents under max_tokens_per_doc
    3. Use top_n to limit results
    Did this fix it?
  3. 3. Handle rate limits and 503

    1. Retry with backoff on 429 and 503
    2. Production Rerank limit is 1,000 requests per minute
    Did this fix it?

Symptoms

Rerank returns 400, 422, 429 or 5xx, or relevant passages seem ignored.

Why it happens

  • Missing or invalid fields (model, query, documents)
  • More than about 1,000 documents in one request (recommended limit)
  • Documents longer than max_tokens_per_doc (default 4096) are truncated
  • Trial Rerank limit of 10 requests per minute
  • Using retired rerank v2 model names

This guide also covers

  • cohere rerank error
  • cohere rerank 422
  • cohere rerank documents limit

Still not working?

If Cohere keeps failing, you're not alone. See every Cohere problem users report or the alternatives people switch to.

Have an app idea? Our small team builds any software: web and mobile apps, games, desktop, AI tools. Price agreed together.

Talk to us
Fixed it? Share this guide

The same problem in other apps

All not loading, error messages & connection guides →

Other Cohere errors