Rerank 4 Fast
Rerank 4 Fast is the throughput-focused tier of the previous reranking generation. It supports multilingual document ranking and remains a listed model.
Model behavior
The Fast and Pro variants use the same 32K context class. The model is accessed through a reranking endpoint and returns relevance ordering rather than generated prose.Oracle documentation for Cohere Rerank 4
How to use it well
Practical guidance. Keep this model in a comparison when an existing search service already meets response-time targets with it. Review whether Rerank 5 Fast improves difficult queries without making the application slower. Use the same candidate lists to isolate the reranker's effect. If you change retrieval and reranking together, an improvement or regression becomes harder to explain.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | API |
| Model identifiers | rerank-v4.0-fast |
| Context window | 32,768 tokens Query and document share the context; reserved tokens also consume space. Rerank context, truncation and input guidance |
| Generation | Previous generation; still listedCohere model catalog |
Related models
- Rerank 5 Fast
Cohere · Rerank
Low-latency search reranker with calibrated relevance scoring and Context Focus.
Launch date not confirmedAPI - Rerank 4 Pro
Cohere · Rerank
Multilingual search reranker for documents and semi-structured records, with a focus on ranking quality.
Launch date not confirmedAPI