Back to Reranking & retrieval

Model comparison

Cohere Rerank 4 Pro vs Cohere Rerank 4 Fast

Compare Cohere Rerank 4 Pro and Cohere Rerank 4 Fast using the same provider-sourced reranking & retrieval rubric. No mystery score and no invented benchmark ranking.

Facts checked September 4, 2026

Quick take

Cohere Rerank 4 Pro

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

Best for

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

Watch out for

One billing unit includes up to 100 documents, not the endpoint's full 10,000-document request limit. Long documents can become several billable chunks.

Cohere Rerank 4 Fast

Cohere's lower-latency Rerank 4 option for high-volume search, recommendations, and RAG retrieval.

Best for

  • High-throughput product search
  • Latency-sensitive RAG applications
  • Teams A/B testing speed and retrieval quality

Watch out for

Fast is a relative product tier, not a latency guarantee. Measure it with your document lengths, region, candidate count, and traffic pattern.

Compare the published facts

Cohere Rerank 4 Pro vs Cohere Rerank 4 Fast

Values use each provider's own published units and limits. A blank means the provider did not publish a directly comparable value in the sources reviewed.

Reranking & retrievalCohere Rerank 4 ProCohere Rerank 4 Fast
Price basisThe provider's billing unit, with the document or token assumptions needed to compare it fairly.$2.50 / 1K searches; up to 100 docs per unit$2 / 1K searches; up to 100 docs per unit
Text capacityHow much query and candidate text the model can consider in one scoring pass.32,768 tokens per query-document pass32,768 tokens per query-document pass
Candidates per requestThe maximum number of possible results accepted in one request, when the provider publishes it.Up to 10,000 documents per requestUp to 10,000 documents per request
Language coveragePublished multilingual support; an exact count is shown only when the provider gives one.Multilingual; exact count not publishedMultilingual; exact count not published
Structured dataWhether the model is documented for JSON, tables, XML, or other data beyond normal prose.Text and semi-structured JSONText and semi-structured JSON
Where it runsHosted API, cloud marketplace, private deployment, or self-hosted weights.Cohere API; enterprise private and cloud optionsCohere API; enterprise private and cloud options

How to choose

Compare the job, not the hype.

Start with the job you need to complete, then validate cost, access, and policy details on your exact provider route.

Cohere Rerank 4 Pro

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.

Cohere Rerank 4 Fast

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.

Frequently asked questions

Cohere Rerank 4 Pro vs Cohere Rerank 4 Fast FAQ

What is the main difference between Cohere Rerank 4 Pro and Cohere Rerank 4 Fast?

Cohere Rerank 4 Pro: Cohere's quality-first reranker for improving search and RAG results across multilingual text and semi-structured JSON. Cohere Rerank 4 Fast: Cohere's lower-latency Rerank 4 option for high-volume search, recommendations, and RAG retrieval.

Should I choose Cohere Rerank 4 Pro or Cohere Rerank 4 Fast?

Consider Cohere Rerank 4 Pro when your priority is High-value enterprise search. Consider Cohere Rerank 4 Fast when your priority is High-throughput product search. Test both with your own data and provider route before committing.

Is this Cohere Rerank 4 Pro vs Cohere Rerank 4 Fast comparison based on Cody benchmarks?

No. This comparison aligns provider-published facts for the Reranking & retrieval category. It does not claim a universal winner or combine incompatible third-party benchmark scores.