All models
Cohere

Cohere Rerank 4 Pro

Cohere's quality-first reranker for improving search and RAG results across multilingual text and semi-structured JSON.

Plain-English overview

What Cohere Rerank 4 Pro actually is

Rerank 4 Pro does not replace a search index. It takes the best candidates from keyword or vector search, reads each one alongside the user's query, and returns a better ordering with relevance scores.

The Pro variant is intended for harder, quality-sensitive searches. Its 32K window can inspect long candidates, while automatic chunking helps with oversized documents. Billing is based on searches, so document count and chunk length matter as much as request count.

Good fit for

  • High-value enterprise search
  • RAG pipelines where retrieval quality matters more than minimum latency
  • Multilingual or JSON-heavy knowledge bases

Category comparison

The facts that matter for reranking models

These are provider-published specifications, not Cody benchmark scores. Follow the linked sources for current limits and endpoint-specific exceptions.

Price basis
$2.50 / 1K searches; up to 100 docs per unitThe provider's billing unit, with the document or token assumptions needed to compare it fairly.
Text capacity
32,768 tokens per query-document passHow much query and candidate text the model can consider in one scoring pass.
Candidates per request
Up to 10,000 documents per requestThe maximum number of possible results accepted in one request, when the provider publishes it.
Language coverage
Multilingual; exact count not publishedPublished multilingual support; an exact count is shown only when the provider gives one.
Structured data
Text and semi-structured JSONWhether the model is documented for JSON, tables, XML, or other data beyond normal prose.
Where it runs
Cohere API; enterprise private and cloud optionsHosted API, cloud marketplace, private deployment, or self-hosted weights.

API and provider access

Where to get Cohere Rerank 4 Pro

Availability

Regions and access stage

Available through Cohere's Rerank API, with private and third-party cloud deployment options available for enterprise use.

Hosted and private-deployment regions depend on the selected Cohere route and enterprise agreement.

Check live availability

Data and training

The route matters.

Cohere enterprise customers can opt out of training; SaaS prompts and generations are generally deleted after 30 days. Approved zero-data-retention accounts and private or third-party deployments offer stronger controls.

This is a concise reading of the cited provider material, not legal advice. A third-party gateway can have different storage, routing, training, and residency terms from the model maker's direct API.

Read the provider policy

Frequently asked questions

Cohere Rerank 4 Pro FAQ

What is Cohere Rerank 4 Pro?

Cohere's quality-first reranker for improving search and RAG results across multilingual text and semi-structured JSON. Rerank 4 Pro does not replace a search index. It takes the best candidates from keyword or vector search, reads each one alongside the user's query, and returns a better ordering with relevance scores.

When was Cohere Rerank 4 Pro released?

Cohere Rerank 4 Pro was released on December 11, 2025 according to the cited provider materials.

Where can I access Cohere Rerank 4 Pro?

Available through Cohere's Rerank API, with private and third-party cloud deployment options available for enterprise use. The access routes listed in this guide are Cohere.

How much does Cohere Rerank 4 Pro cost?

$2.50 / 1,000 searches. One search unit covers one query against up to 100 documents. Documents or chunks above the pricing token allowance can count as additional documents.

Where is Cohere Rerank 4 Pro available?

Available through Cohere's Rerank API, with private and third-party cloud deployment options available for enterprise use. Hosted and private-deployment regions depend on the selected Cohere route and enterprise agreement.

Is my Cohere Rerank 4 Pro API data used for training?

Cohere enterprise customers can opt out of training; SaaS prompts and generations are generally deleted after 30 days. Approved zero-data-retention accounts and private or third-party deployments offer stronger controls. The policy belongs to the provider route and account terms, so verify it again before production use.